Dell EMC’s PowerEdge R550 is aimed squarely at SMBs that have processing power, storage capacity and expansion potential on their shopping list and a tight grip on the purse strings. This 2U rack server looks to deliver on all counts; along with support for dual Xeon Scalable CPUs, it offers a good range of storage features plus plenty of room to grow, and yet doesn’t cost much more that an equivalent 1U model.

It bears a number of similarities to the 1U PowerEdge R450 as it too is offered with a choice of seven Xeon Scalable Gen 3 CPUs. It has the same maximum 165W TDP thermal restriction but this still allows it to support up to 24-core Xeon Scalable Gold 5318Y CPUs - our system was supplied with one of these in residence.

Businesses with storage-centric workloads will find plenty to like here. The R550 supports SAS3 and SATA devices and has room up front for 8 LFF HDDs or 16 SSF HDDs and SSDs, and with 20TB SATA HDDs already available from Dell EMC, the R550 can present a very usable maximum capacity of 160TB.

Dell EMC PowerEdge R550 review: Internal design

The R550 presents a well-designed interior with easy access to all key components for upgrade and maintenance operations. Unlike Dell EMC’s higher-end rack servers such as the R650 and R750, it doesn’t have a T-shaped motherboard and keeps its pair of PSU bays over on the left side.

Entry-level systems can get away with dual 600W redundant PSU but stepping up to dual Gold CPUs will require 800W models. Dell EMC also offers 1,100W versions, although we can’t see many specifications requiring these – especially as the R550 doesn’t support GPU cards.

Cooling is handled efficiently by five hot-plug fans arranged in a removable tray in front of the motherboard. With one CPU in residence, we only needed the standard single rotor variety, whereas dual CPUs will require the high-performance Silver-Grade fan modules fitted.

Our Gold CPU was covered by a sturdy finned passive heatsink and the two sockets are partnered by 16 DIMM slots, with the price for our system including 128GB of memory spread across eight 16GB modules. The maximum memory is 1TB which is perfectly adequate for most businesses - in fact, Dell EMC informed us that it found few customers buy more than this.

Dell EMC PowerEdge R550 review: Storage and expansion

Our server was supplied with 8 LFF hot-swap drive bays across the front and you can specify either 8 or 16 SFF bays. For RAID, you can start with the embedded PERC S150 controller, which provides software-managed arrays, but the price we’ve shown includes Dell EMC’s top-dog PERC H755 Front SAS card.

Along with support for all popular RAID arrays types, it sports a big 8GB of DDR4 cache memory and comes with a battery backup pack. A feature which makes for a very tidy interior is that the RAID card is mounted above the drive bays, plugs directly into the backplane and connects to the motherboard with one cable.

Unlike the R450, the R550 supports Dell EMC’s latest BOSS (boot optimised storage solution) S2 card. This presents dual M.2 SATA SSDs in removable carriers at the rear and is ideal for loading an OS or hypervisor on, saving all the front bays for data storage.

At the rear are four PCI-E slots with each accepting a vertically mounted adapter card. Clearly, there’s plenty of room for more components with our single CPU activating one PCI-E 4 slot plus the PCI-E 3 version alongside, with the other two PCI-E 4 slots enabled when a second CPU is installed.

You probably won’t need the standard slots for network expansion though, as sitting in the middle is an OCP 3 mezzanine edge slot. The price we’ve shown includes a Broadcom dual 10GbE SFP+ card with other choices comprising multi-port Gigabit, 10GbE and 25GbE cards from Broadcom, Intel and Marvel.

Dell EMC PowerEdge R550 review: Management

The server’s embedded iDRAC9 controller sets the standard for remote management with its web console offering a wealth of information on all hardware components, system performance and thermal values. It provides direct access to BIOS or storage configurations and we’ve included a Datacenter license in the price, which activates everything the iDRAC9 has to offer, including remote console and virtual media services plus telemetry streaming for predictive analytics.

Dell-centric businesses will like the OpenManage Enterprise (OME) software which can monitor and report on PowerEdge servers, storage arrays and virtualization hosts. We run it in the lab as a Hyper-V VM and use it to check on system utilization, firmware compliance and warranty status.

We’ve also included an OME Advanced license for the R550 so it can send power usage statistics to the OME Power Manager plug-in. This adds group power consumption graphs to the OME dashboard page plus individual statistics for each server, with data being retained for up to a year.

Adding another plug-in to the OME host allows it to function as a collector for Dell EMC’s CloudIQ cloud monitoring service so it can send telemetry and alert logs to it for selected groups of servers, storage arrays and switches. CloudIQ requires all monitored systems to have a valid ProSupport contract and for the R550, the portal presented a proactive health score along with detailed lists of all hardware components plus a 24-hour performance view for all major elements.

Dell EMC PowerEdge R550 review: Verdict

The PowerEdge R550 is a fine choice for budget-conscious SMBs that want a versatile and powerful Xeon Scalable rack server. It has the storage density to satisfy data hungry applications and combines it with plenty of expansion potential and classy remote management features.

Dell EMC PowerEdge R550 specifications (as reviewed)

As one of the first Gen3 Xeon Scalable servers to market, Dell EMC’s PowerEdge R750 set a very high standard and impressed us mightily when we exclusively reviewed it at its launch. It packs a truly remarkable specification into its 2U of rack height and offers a huge expansion potential as well.

The only issue with the R750 is it could be overkill for businesses that have more modest rack server requirements, as they could end up paying for features their workloads may never use. Not everyone needs core-heavy Xeon Scalable Platinum CPUs, a massive 8TB of system memory, Intel PMEM 200 modules or liquid-cooling, after all.

Enter the PowerEdge R750xs, which is designed for businesses that don’t need such a high-end specification, taking a lot of the goodness from the R750 and offering it as a more cost-optimised version. Along with large server scale-out projects, the R750xs targets more general customer workloads such as databases, virtualization, VDI and cloud services.

Dell EMC PowerEdge R750xs review: Hardware features

If you think the R750xs is a ‘hobbled’ version of the R750, you’d be very wrong. For starters, it can handle all Gen3 Xeon Scalable Gold CPUs right up to the 32-core 6338/6338N models and the price for our system includes a very serviceable pair of 20-core 2.3GHz Xeon Scalable Silver 4316 CPUs.

Dell EMC found many of its customers aren’t buying more than 1TB of memory, so that’s what the R750xs will support and our system came decked out with a healthy 512GB of DDR4 RDIMM memory spread across all 16 of its DIMM slots. Only those with very specific workloads and deep pockets would likely specify Intel Optane Persistent Memory, so support for these has been removed from the R750xs as well.

Further research found that only a small number of customers were buying more than 12 drives, so storage arrangements see some changes too. A quick glance at the front shows the R750xs on review retains the same 8+8+8 drive bay layout as the R750 - but options extend to 16 SFF SAS/SATA devices, with the third bay set aside for those that want NVMe SSDs.

If you need more room to maneuver, you can also specify either 8 or 12 LFF front drive bays, and an optional rear two-bay cage is available which supports HDDs and NVMe SSDs. This is in contrast to the R750, which supports two rear cages. Our server also has Dell EMC’s BOSS (boot optimized storage solution) S2 card which presents its two M.2 slots at the rear, with 480GB SATA SSDs fitted in removable hot-swap carriers for easy access.

Dell EMC PowerEdge R750xs review: Chassis design

The R750xs chassis is slightly shorter than its big brother and has a different internal design. Whereas the R750 employs a T-shaped motherboard which allows its PSUs to be placed on each side of the chassis for improved cooling, the R750xs keeps the older design, with the dual PSUs installed in one rear bay to the left.

Cooling isn’t a problem though, as the server is available with five standard hot-plug fan modules which can handle CPUs up to a 150W TDP. Higher specified systems require either Silver or Gold high-performance fans and there’s a sixth bay which will be fitted with an extra fan module if the rear drive cage is installed.

Expansion doesn’t see any major restrictions as the R750xs supports up to six PCI-E Gen 4 half-length, half-height cards with an OCP 3 edge connector lurking beneath the central 2-slot riser. The price of our system includes an Intel X710 dual-port 10GbE SFP+ OCP 3 card with Dell EMC also offering 10/25GbE mezzanine cards.

Dell EMC PowerEdge R750xs review: Management and CloudIQ

Remote server management is as excellent as we’ve come to expect from Dell EMC, and the server’s iDRAC9 controller offers a feast of features presented in its tidy web console. You can keep a close eye on all aspects of server operations, view performance, remotely configure RAID arrays and directly access its BIOS menus.

Dell EMC’s OpenManage Enterprise (OME) server provides a web console for managing all your servers, with the Power Manager plug-in adding system consumption and thermal monitoring. The free CloudIQ cloud hosted service goes even further by using machine learning to deliver server telemetry and predictive analytics for your entire Dell EMC infrastructure.

To use CloudIQ, we added its plug-in to our OME on-site installation, generated an access code and PIN and selected the OME device groups we wanted monitored. OME functions as a collector for CloudIQ, sending telemetry and alert logs to it every 15 minutes, and the only criteria for servers is they must have a valid support contract.

The CloudIQ portal opens with an overview of all monitored servers, storage arrays and switches, shows system health status and alerts while the cybersecurity feature identifies potential security violations. CloudIQ performs predictive analysis on storage pools and clusters and provides historical trending and timely warnings as to when capacities are likely to reach their limits.

Plenty of information was provided for our R750xs review server including a proactive health score, detailed lists of all hardware components and a 24-hour performance view for CPU and memory utilisation, power consumption and temperatures. The system performance page presents views of system level performance for all monitored devices which can be customized with filters while the reports section allowed us to create custom dashboards showing graphs and tables for selected server metrics.

Dell EMC PowerEdge R750xs review: Verdict

Businesses that want an affordable 2U rack server without all the extra bells and whistles will find Dell EMC’s PowerEdge R750xs fits the bill nicely. With support extending to all Gen 3 Xeon Scalable Gold CPUs and up to 1TB of memory, it’s capable of handling virtually any general-purpose workload, storage and expansion potential are both equally good and the system on review offers a fine specification for the price.

Dell EMC PowerEdge R750xs specifications (as reviewed)

Multi-node servers are perfect for data centers demanding ever greater compute densities without an increase in real estate and Dell EMC's PowerEdge C-Series of rack servers have traditionally offered an appealing solution. Built around its trusty C6400 chassis, they support up to four independent compute nodes and the PowerEdge C6520 brings Intel's 3rd Generation 'Ice Lake' Xeon Scalable CPUs to the table.

The C6520 nodes are the dual-socket (2P) variety and support every Xeon Scalable model right up to Intel's mighty 40-core 2.3GHz Platinum 8380. Processing density isn't as high as Dell EMC's AMD-based PowerEdge C6525 but Intel shops will appreciate being able to cram up to 320 physical and 640 logical cores into only 2U of rack space.

The C6520 targets a wide range of enterprise workloads such as HPC, massive parallel computing, financial modelling, high-frequency trading and, of course, server consolidation. For those that require a more scalable performance, reduced operating costs and lower cooling requirements, the C6520 can be deployed as a rack dense solution for 1P workloads.

Dell EMC PowerEdge C6520 review: C6400 chassis

The C6400 chassis has seen an extensive tour of duty: it's the same as employed by the PowerEdge C6525 and also the 14th Gen PowerEdge C6420 which was first launched back in 2017. In reality, there isn't much that needs changing as its primary functions are to provide power to the nodes along with storage backplane services.

The chassis uses a central mid-plane management board to distribute power to the nodes and Dell EMC offers a wide range of hot-plug PSUs. The review server was supplied with dual 2,000W PSUs, but 1,600W, 2,400W and 2,600W versions are also available.

Our system had 24 SFF drive bays (a 12-bay LFF model is also available) and each node gets an equal share of drives which, in the standard configuration, can be all SAS3/SATA, with the first two bays of each group also supporting NVMe SSDs. Note, however, that this is a legacy mixed-mode backplane and only offers PCI-e 3 speeds, whereas the new all-NVMe backplane delivers PCI-e 4 performance.

The air-cooled version of this chassis uses four cold–swap fans but has a number of thermal restrictions, as most Xeon Scalable CPUs with TDPs of 235W and above are not supported in DP configurations. However, Dell EMC offers a DLC (direct liquid cooling) option which is up to four times more efficient than air-cooling and removes all thermal restrictions.

Dell EMC PowerEdge C6520 review: Nodes

Our review system was supplied with a single node equipped with dual 2.1GHz 36-core Xeon Scalable Platinum 8352V CPUs which are optimised for virtualization duties. Their 195W TDP puts them on the right side of any thermal restrictions and CPU 1 was also fitted with the extended heatsink kit which provides an extra 5℃ of leeway for continuous operating temperatures.

The CPU sockets are flanked on each side by four DIMM slots and ours were all populated by 16GB 3,200MHz DDR4 RDIMM modules. Memory capacity can be boosted to 1TB using RDIMMs and 2TB with LR-DIMMs.

Expansion potential is good, as the nodes have two PCI-e Gen 4 slots, although the extended heatsink loses one of them. No matter; underneath this is an OCP 3.0 mezzanine card edge slot which accepts open-standard OCP 3.0 network cards from other partners, as well as Dell EMC's own Gigabit, 10GbE, 25GbE and 100GbE adapters.

The node's embedded S150 controller supports SATA devices and offers software-managed stripes, mirrors and RAID5 arrays but SAS3 support will require an additional PERC controller to be fitted. The C6520 supports a good range, with the top dog PERC H745 RAID card supporting SATA/SAS devices and offering plenty of RAID array options.

For workloads that don't require much storage, you can specify a C6400 chassis without a backplane, which will improve your thermal outlook as airflow is less restricted. The nodes have a dedicated slot at the rear which supports a BOSS-S1 (boot optimized storage subsystem) dual M.2 SATA adapter, which provides mirrored storage for running an OS.

Dell EMC PowerEdge C6520 review: Remote management

The C6520 nodes come with a single Gigabit LOM (LAN on motherboard) card, which slots in underneath the right expansion riser and can be shared with the embedded iDRAC9 controller. Remote management features are tops, and the iDRAC9 web console provides a wealth of information about hardware health plus operational status - and unlike Dell EMC's AMD-powered servers, you also get the System page performance tab, which provides graphs of CPU, memory and I/O utilization.

Multiple servers and nodes can be centrally managed using Dell EMC's OpenManage Enterprise (OME). We run the latest v3.4 in the lab as a Hyper-V VM and used its discovery service to add our C6520 node, after which we could view its vital signs, remotely control it and employ OME's firmware services to ensure it was running the latest version.

For more control, the OME Power Manager plug-in is a worthy addition as it can be used to pull up historical views of power consumption and thermal values going back up to a year. Even better, it can enforce per-rack power cap and thermal event-based policies while its EPR (emergency power reduction) feature can apply a low power mode to all group members to control consumption and if necessary, gracefully shut them all down.

Dell EMC PowerEdge C6520 review: Verdict

The PowerEdge C6520 may not be able to match the monstrous core density offered by the AMD-based PowerEdge C6525 but it still packs an impressive amount of Gen3 Xeon Scalable power into a rack-friendly 2U chassis. Available in air and liquid cooled configurations, the nodes are well designed with good expansion potential, while remote management features are simply the best - a great choice for high performance or hyperscale workloads and Intel houses with big server consolidation plans.

Dell EMC PowerEdge C6520 specifications (as reviewed)

The server market has seen some big changes recently. but Dell EMC looks set to shake it up even further with the largest product launch in its history. The 15th-generation PowerEdge family comprises 17 new and refreshed models and in this hands-on review, we look at the latest version of its trusty 2U data centre workhorse.

As expected, the PowerEdge R750 supports Intel’s 3rd Generation ‘Ice Lake’ Xeon Scalable CPUs which finally give AMD’s well-established EPYC chips some serious competition. Employing a 10nm architecture, they range from 8 to 40 cores and up to 64 PCI-E 4 lanes per socket, support DDR4 memory speeds up to 3,200MHz and claim an average 46% performance improvement over the previous generation for common data centre workloads.

Three versions of the R750 are available, but we’re focusing on the standard general-purpose model. The R750xa is built for GPU-based workloads and claims the highest density of GPUs in a 2U rack server while the R750xs is a cost-optimized version aimed at specific customer workloads.

The R750 presents a radical redesign over the R740 with a sharper focus on internal cooling. Storage capacities stay the same, with a maximum of 28 SFF bays available, but NVMe SSDs get some serious attention and expansion potential receives a massive boost.

Dell EMC PowerEdge R750 review: Internal layout

Internally, the R750 has a similar layout to the AMD-based PowerEdge R7525 as its T-shaped motherboard allows the PSUs to be placed on each side of the chassis. This improves air flow through the chassis so there are no hot spots and on the 1U R650 model, it creates extra exhaust lanes to help cool the CPUs.

Dell EMC’s multi-vector cooling creates streamlined pathways so air can be directed where it’s needed most using automation or user-defined profiles. These features aim to reduce the need for expensive liquid-cooling but in rack dense environments where heat dissipation is a major requirement, you can specify direct liquid cooling (DLC) kits that feature proprietary leak detection systems and shutdown services.

A bank of six hot-plug cooling fans are arranged in front of the motherboard and our server came with the Silver series modules. If you want the core-dense, high TDP Xeon Scalable Platinum 8300 CPUs, you’ll need to specify the higher performing Gold fans.

General internal design is very tidy with the two 28-core 2GHz Xeon Scalable Gold 6330 CPUs in our system laid out neatly across the chassis, mounted with solid passive heatsinks and flanked by 16 DIMM slots per socket. The price of our server includes 1TB of DDR4 RDIMM currently upgradeable to 2TB, while the maximum 8TB can be achieved with a full house of 256GB LRDIMMs or sixteen of Intel’s latest Optane 200 persistent memory modules.

Dell EMC PowerEdge R750 review: Storage features

The R750 offers an extensive range of storage options that start with a 12-bay LFF hot-swap backplane for SAS3/SATA devices. We have the model with the 16-bay SAS/SATA SFF backplane and Dell EMC provided eight 2.4TB SAS3 SFF HDDs which we’ve included in the price shown.

For NVMe SSDs, you have a number of options as in our system, there’s room to the right for an extra 8-bay backplane. You can go for a full house of NVMe SSDs with a 24-bay backplane but this doesn’t support SAS/SATA devices. It’ll cost you a PCI-E riser but you can add a 2-bay or 4-bay enclosure at the back and choose from SAS/SATA or NVMe SSD versions.

RAID sees big improvements as the new PERC11 controllers have been moved on to dedicated slots on the storage backplane. We have the standard PERC H755 front SAS model and the PERC H755N model finally brings hardware RAID to NVMe devices as this PCI-E 4 adapter has the bandwidth to make this a reality.

The BOSS (boot optimized storage solution) card gets a big makeover as well. Instead of being buried inside the server, the new S2 version presents its two M.2 slots at the rear where SATA SSDs are fitted in removable hot-swap carriers.

Dell EMC PowerEdge R750 review: Expansion and management

Expansion potential is outstanding as the revamped internal design provides sufficient room for four dual-slot PCI-E 4 risers. A range of risers are available and include a Snap I/O version that currently supports Infiniband adapters and allows them to have a x8 connection to each CPU for reduced latency.

What’s more, underneath the central riser is an OCP 3.0 mezzanine edge slot. This replaces the older proprietary LOM surface port so the R750 can use open-standard network cards from other partners as well as Dell’s own offerings.

The iDRAC9 controller doesn’t see any major improvements but none are needed - it already provides the best remote management features in town. Its web console presents a ton of information on server operations plus hardware health, with the Datacenter license adding advanced services such as streaming telemetry for predictive analytics.

We run Dell EMC’s OpenManage Enterprise (OME) server in the lab as a Hyper-V VM allowing us to monitor and configure all our PowerEdge servers from one web console. To test the Power Manager plug-in, we imported an iDRAC9 Enterprise Advanced license to the R750, added it to our monitored device group and used the OME home page to view historical graphs of group power consumption ranging from 6 hours to a year, check on thermals and see which systems were using the most power.

The plug-in will prove invaluable to data centres as it supports up to 8,000 systems and can enforce per-rack power cap and thermal event-based policies to automatically throttle servers back when they’re getting toasty. And if the situation becomes critical, the EPR (emergency power reduction) feature can push out an instant mass policy to selected groups forcing all members to drop to a low power mode or be gracefully shut down.

Dell EMC PowerEdge R750 review: Verdict

As one of the first Gen3 Xeon Scalable servers to market, the PowerEdge R750 sets a very high standard for the rest to follow. The redesigned chassis delivers improved cooling, a huge expansion potential plus plenty of great storage features, and combining these with top-notch remote management services makes this 2U rack server ideally suited to a wide range of common data centre duties.

Dell EMC PowerEdge R750 specifications (as reviewed)

Most servers can be upgraded and expanded as your needs grow, but HPE’s ProLiant DL180 Gen10 has more scope for growth than most. That’s thanks to a second CPU socket that allows you to move up to multiprocessor computing as and when you’re ready.

The entry price, meanwhile, is low, with lightweight specifications starting at just £986. This modest sum gets you out of the starting blocks with a six-core Xeon Scalable Bronze 3204 CPU, 16GB of DDR4 Smart Memory and eight LFF drive bays, cabled to the embedded HPE S100i software-based RAID controller.

Move up to the £1,345 model on test and you get an eight-core Silver 4208 CPU, a full house of 12 LFF drive bays and HPE’s Smart Array P408i SATA/SAS3 RAID card with 2GB of cache and battery backup unit. Clearly, this is a configuration geared up for storage-centric workloads, but SFF fans will be pleased to know that the basic eight-bay model can be upgraded to 16 or 24 bays, with the option to add two more at the rear.

The motherboard has both internal USB 3 and microSD card slots, which you can use to boot into an embedded hypervisor. To avoid wasting storage bays on boot media, HPE’s optional PCIe Enablement board supports a pair of mirrored M.2 SSDs to house your operating system, and is cabled to dedicated SATA interfaces on the motherboard.

Remote management functions are excellent. The server comes with HPE’s iLO5 chip and a Standard licence, which provides a stiff set of platform security features such as “silicon root of trust” to protect the server’s firmware and a web interface packed with valuable information about critical components. Be aware, though, that if you want to enable power metering, OS remote control and virtual media services, you’ll need to upgrade to the Advanced licence.

To try out the management features, we added the server’s iLO 5 details to the lab’s virtualised Hyper-V HPE OneView 5.5 app. We were then immediately able to monitor the health of all our ProLiant servers and gather inventory, and if you’ve shelled out for the Advanced iLO 5 licence then OneView can also provide data on system utilisation, thermals and power usage. However, in order to provision new servers and deploy automated firmware updates you’ll need yet another upgrade, this time to the OneView Advanced licence.

Internally, the server presents a tidy layout, with the Smart Array P408i RAID card residing in a slot at the rear of the motherboard and cabled directly to the drive backplane. The server includes a riser with three PCIe slots, held in place by a locking tab that can be easily released with a flick of the wrist. If you install a second CPU, you can add another three-slot PCIe riser card, which fits next to the PSU bay.

Underneath the riser is a media module slot; HPE offers dual-port Gigabit Ethernet and 10GbE adapters that can be used to supplement the server’s dual embedded Gigabit ports. The CPU sockets are flanked on each side by banks of four DIMM slots. The base 16GB of DDR4 in the review system can be boosted to a maximum of 1TB, and installing the second CPU activates its attendant DIMM slots, allowing memory capacity to be doubled to 2TB.

Expanding businesses searching for a powerful Xeon Scalable server with a big upgrade path will find the ProLiant DL180 Gen10 a great choice. The price above doesn’t include any storage, and you may well want to factor in the cost of upgrading the iLO5 controller too, but base systems are very affordable and provide enormous expansion potential.

HPE ProLiant DL180 Gen10 specifications

The blue chips have been slow to embrace AMD’s EPYC server processors, but for Taiwanese server specialist Tyan it’s been full steam ahead: the company now has a portfolio of over 30 motherboard designs supporting the latest core-heavy Gen2 EPYC 7002 CPUs.

From this line-up, Broadberry has chosen the Transport HX TS75-B8252 barebones system as the basis for its CyberServe AM2-B8252 server – and the end product really shows off the potential of the platform. The price above includes not one but a pair of mighty 32-core 2.4GHz AMD EPYC 7532 CPUs, and a whopping terabyte of 3,200MHz DDR4 memory. There’s 80TB of storage too, in the form of eight 10TB SATA hard disks – and to hammer the point home, Broadberry includes a dual-port 100GbE card.

Perhaps the most remarkable part is that this isn’t even a maxed-out configuration. The AM2-B8252’s 32 DIMM slots can support up to 8TB of memory, and all expansion slots are of the ultra-fast PCI-E 4 variety, opening up no end of possibilities.

Cracking open the lid reveals that Tyan has taken a leaf out of Dell EMC’s book when it comes to design, as the layout is very similar to that of the PowerEdge R7525. Splitting the PSUs across the rear of the chassis makes the CyberServe rack PDU friendly, and allows for a balanced air flow, so there isn’t a single hotspot on one side of the server.

The twin CPU sockets are flanked on each side by those plentiful DIMM slots, with cooling handled by a bank of six hot-plug fans arranged along the front of the chassis. Considering the 200W TDP of the CPUs, these are impressively quiet: we measured a modest overall noise level of 61.5dB from one metre in front.

Storage-wise, alongside the eight hard disk bays, you’ll find empty hot-plug bays for four NVMe SSDs. All bays use a simple hinged locking lever, with colour-coding to help you identify the SSD bays at a glance, and there’s also a pair of M.2 NVMe SSD slots, supporting all card sizes up to 110mm. With all these storage options to hand, it’s slightly surprising to note that the server doesn’t feature an embedded RAID controller – but it’s no problem to install your own, and BroadBerry offers an eight-port LSI SAS3 option for £495.

Once that’s in place, you’re still left with plenty of expansion options. The CyberServe employs four riser cages to present a total of four PCI-E 4 x16 slots and one PCI-E 4 x8 connector. If you require even more expansion options, there’s also a variant of the motherboard available that provides nine x8 slots.

We’ve no complaints about the networking provision: the Broadcom 100GbE card comes in addition to dual embedded 10Gbase-T ports, and you can easily increase the port count as the server offers an OCP (Open Compute Project) 3.0 Gen4 slot, with a smart tool-free design allowing you to slot cards in from the rear without having to open the server up.

If there’s one area where Tyan’s board doesn’t quite measure up to the competition, it’s remote management; the AST2500 controller simply can’t match HPE’s iLO5 or Dell’s iDRAC9 for features. Nevertheless, the web console provides all the important details about critical components, with a long list of sensor graphs that can be drilled down into for more detail. There’s also a PEF (platform event filter) service that can be used to create multiple policies linking events and errors with email alerts, and great access security courtesy of a system firewall and support for LDAP, RADIUS and Active Directory authentication. Full OS remote control and virtual media services as standard too, rather than being held back as optional upgrades.

The CyberServe AM2-B8525 is clearly overkill for a small business role, but SMEs and enterprises seeking an uncompromising rack server at a great price will find it hard to ignore. Not only is it an astonishingly capable server as supplied, it has plenty of room for further growth as your requirements evolve.

Broadberry CyberServe AM2-B8252 specifications

AMD’s latest core-dense server CPUs are a huge challenge to Intel, and its three new 7Fx2 chips promise speeds, cache capacities and prices that their Xeon equivalents just can’t match.

Broadberry’s CyberServe R272-Z31 showcases their potential. It’s built around the EPYC 7F72 model, which features 24 cores, a 3.2GHz base clock speed that’s higher than the Xeon Scalable Platinum 8268 and a 192MB L3 cache that’s more than five times the size of its rival’s. It also supports up to 2TB of 3,200MHz RDIMM DDR4 memory, and a huge 128 PCIe 4 lanes.

While this is undeniably a high-end server, the R272-Z31 is good value for virtualisation, as it doesn’t attract VMware’s extra licensing charge for CPUs with more than 32 physical cores. Similar benefits apply if you’re running SQL databases with core-based licensing models.

Storage options are cost-effective too: the price above doesn’t include drives but, unlike some vendors, Broadberry leaves you free to source and fit your own, with no impact on the main warranty. An impressive 24 hot-swap SFF bays at the front with two more at the rear should be adequate for hyper-convergence duties, and all carriers are included in the price.

Architecturally, the CyberServe is built on a Gigabyte R272-Z31 2U barebones rack system. It’s a safe choice, as Gigabyte offers more EPYC 7002 server platforms than all the blue chips put together. It also means you get the benefit of Gigabyte’s modularised drive backplane, and a good choice of exchangeable expander boards. A total of one PCIe 3 and five PCIe 4 slots are on offer, and our system came with a single SAS 3/SATA backplane expander with all 26 bays connected to it and routed to a MegaRAID 9361-8i PCIe 3 card. If you want more storage performance, you can pay an extra £200 for the MegaRAID 9560-8i, the industry’s first PCIe 4 RAID card.

The Gigabyte design also includes a powerful cooling system, designed to cope with the high 240W TDP of the EPYC 7F72 CPU. By default, the four main fans in front of the motherboard spin at a huge 6,900rpm: this certainly keeps the CPU cool, but it’s very noisy, with our iPad recording sound levels of 73.5dB at one metre in front. It also drains a lot of power: after some experimentation in the server’s web-based remote management console, we settled on a profile that drops the fans to a calmer 4,200rpm when the CPU temperature is below 30°C, and only ramps them up when it goes above this.

On that note, the AST2500 web console won’t be worrying the likes of Dell’s iDRAC9, but it provides plenty of information on critical components with views of all hardware sensors. It includes OS remote control and virtual media services as standard too, rather than optional upgrades.

You can additionally enable the built-in Gigabyte Server Management utility, which presents a web-based status dashboard for remotely monitored systems. The latest version 2.03 includes power consumption statistics for EPYC 7002 servers, plus (again) detailed views of installed hardware and sensors, fan speed graphs and tools for remote firmware upgrades.

The CyberServe R272-Z31 is a very impressive server. With AMD’s powerful EPYC 7F72 at the helm, and a generous 256GB of RAM included in the price, it’s geared up for heavy-duty processing. At the same time, it lets you save money on per-core licensing and storage provision, making it an excellent all-round proposition that current Xeon Scalable systems simply can’t compete with.

Broadberry CyberServe R272-Z31 specifications

Designed to munch through demanding tasks such as artificial intelligence, machine learning and database analytics, Dell EMC's PowerEdge R840 packs a mighty package into a slim-line 2U rack chassis. Its impressive hardware portfolio aims to get the most out of your rack infrastructure as it supports quad Xeon Scalable CPUs and a whopping 6TB of DDR4 memory.

There's more, as it supports up to twelve high-performance NVDIMM-N modules, has six PCI-Express (PCIe) expansion slots and can handle two double-width GPUs or full-height FPGAs (field programmable logic gates). Storage is high on its agenda too, as there's room for a total of 26 SFF drives and all 24 front bays can be populated with NVMe SSDs.

It's a very different beast to its predecessor, the R830. Along with a radically revamped internal design, the R840 offers twice the memory capacity, a 27% increase in CPU cores and a 62.5% boost in drive bays.

Dell EMC PowerEdge R840: Design

Cracking the lid reveals a new internal design as Dell EMC has done away with the optional PEM (processor expansion module). The R840 uses the same motherboard as the equally new PowerEdge R940xa which presents all four CPU sockets.

Good thermal design allows the R840 to support all Xeon Scalable models right up to the 28-core Platinums. The four sockets are flanked by 48 DIMM slots and to realize the maximum 6TB capacity, you'll need to specify 'M' designated Xeon Scalable CPUs which support 1.5TB per socket.

Our system is 6TB-capable as it came with a tasty quartet of 18-core 2.3GHz Gold 6140M CPUs. The price we've shown also includes a healthy 384GB of DDR4 memory with the 6140M CPUs supporting their highest 2,666MHz speed.

Cooling is handled by a bank of six fan modules arranged behind the drive backplane. These are all hot-pluggable and for easier access to the backplane, the whole fan tray can be easily released and removed.

Dell EMC PowerEdge R840: Storage versatility

The R840 has been designed to offer maximum storage flexibility where base units start with an 8-drive SAS/SATA backplane. This can be cabled directly to the motherboard's pair of SATA ports or to a PERC RAID card to bring SAS into play.

We have the SAS3 expander board and 24-drive active backplane which also has six dual-port PCIe connectors. These are cabled to the corresponding motherboard ports and the server's universal drive bays allow various mixes of SAS/SATA drives and NVMe SSDs to be implemented.

The 24 NVMe option requires four CPUs plus a special backplane and supports dual GPU configurations. You can add a dual-drive SAS/SATA cage at the rear - although this negates GPU support and also reduces the number of available PCIe slots.

RAID options start with the embedded PERC S140 chip which supports SATA drives plus stripes, mirrors and RAID5 arrays. Dell EMC offers a good choice of PERC RAID controllers and our system was supplied with an H740P card which occupies one vertical PCIe slot and comes with 8GB of battery-protected NVRAM cache.

Dell EMC PowerEdge R840: Remote management

Remote management doesn't get any better: the R840 sports Dell EMC's latest iDRAC9 chip along with a dedicated management port. It presents a smart HTML5 web console packed to the gills with valuable information about all critical hardware components.

Supply chain security can be assured as firmware packages are cryptographically signed and storage can be safely wiped using the iDRAC9 System Erase feature. The System Lockdown mode also stops users changing any server configuration settings.

The OpenManage Essentials software is due to be replaced by OpenManage Enterprise. We run the latter as a Hyper-V VM in the lab where it presents plenty of management features and operational data but it still hasn't been updated beyond the Tech Release that was launched a year ago.

Data centres will approve of the optional Quick Sync 2 module. Pressing the button under the system ID LED on the left rack ear activated Bluetooth allowing us to connect our iPad to the server via the OpenManage Mobile (OMM) iOS app and view all server information, alerts and health status.

Dell EMC PowerEdge R840: Power and expansion

Our server was supplied with dual, redundant 1100W Platinum hot-plug PSUs and other options include 750W, 1600W, 2000W and 2400W PSUs. Choose your CPUs carefully if power usage is a concern as the top-end models can be thirsty.

The 6140M CPUs in our system have a 140W TDP and we measured the R840 drawing 147W with Windows Server 2016 in idle but peaking at 810W with all 72 cores under maximum load. The R940 we exclusively reviewed recently had four 105W TDP Gold 5122 CPUs which pulled 145W in idle but peaked at a much lower 510W.

GPU cards are installed in the PCIe risers and the air shroud has all required power cables ready and waiting. If you fit these, you'll need a minimum of dual 1100W PSUs, which must be set to non-redundant mode.

A dedicated slot at the rear is provided for Dell EMC network daughter cards (NDCs). We have the basic quad Gigabit module but other options include dual 10GBaseT/1GbE, quad 10GbE and dual 25GbE NDCs.

Dell EMC PowerEdge R840: Verdict

It may have a low profile rack chassis but everything else about the PowerEdge R840 is big, making it a great choice for enterprises that want to maximize their data centre real estate. It offers high-performance 4P Xeon Scalable processing out of the box, has a big 6TB memory capacity and delivers the highest NVMe SSD density for a 2U rack server.

Verdict

The PowerEdge R840 is a rack-dense 4P Xeon Scalable powerhouse with a big memory capacity and a highly versatile storage arrangement

Chassis: 2U rack

CPU: 4 x 2.3GHz 18-core Xeon Gold 6140M

Memory: 384GB 2,666MHz ECC DDR4 RDIMM (max 6TB)

Storage bays: 24 x SFF hot-swap (max 26)

RAID: Dell PERC H740P SAS3 with 8GB NVRAM cache/BBU

Storage included: 2 x 900GB SAS3 15K SFF HDD

Array support: RAID0, 1, 10, 5, 6, 50, 60

Network: 4 x Gigabit daughter card

Expansion: 6 x PCI-e 3.0 slots

Power: 2 x 1100W Platinum hot-swap PSUs

Cooling: 6 x hot-plug fans

Management: Dell iDRAC9 Enterprise, Quick Sync 2

Warranty: 3Yr ProSupport Plus and 4hr Mission Critical

HPE steals a march on the competition as its ProLiant DL560 Gen10 is one of the first 2U-high 4P Xeon Scalable rack servers to market. Targeting SMEs and enterprises that want the best rack processing density, the DL560 Gen10 is designed to be highly scalable, allowing you to start small and grow as demand increases.

Base systems support two Xeon Scalable CPUs teamed up with 24 DIMM memory slots. When the time is right to step up to 4P processing, just add HPE's mezzanine board kit which delivers two more CPU sockets and another 24 DIMM slots.

A 4P system supports 6TB of DDR4 SmartMemory and can handle up to 24 of HPE's latest 16GB Persistent Memory NVDIMMs. These target high-demand apps such as analytics or databases and combine 16GB of high-performance DRAM plus 16GB of flash mounted on a standard DIMM module and protected by HPE's Smart Storage Battery.

The DL560 Gen10 is certified for all 4P-capable Xeon Scalable CPUs up to 205W TDPs so you can choose from low-cost Xeon 5100 models and scale right up to 28-core Platinums. HPE has also announced support for Intel's latest Gold 6143 and Platinum 8165 high-performance CPUs.

HPE ProLiant DL560 Gen10: Specs and internal design

On review is the entry 840369-B21 model, which gets you out of the starting blocks with a pair of 14-core 2.2GHz Intel Xeon Gold 5120 CPUs partnered by 32GB of DDR4 memory. Remember that the three members of the Xeon 5100 family support memory speeds up to 2,400MHz so to get the higher 2,666MHz speed, you need to specify Gold 6100 or Platinum 8100 CPUs.

Internally, the server exudes a classy built quality and its tidy design affords quick access to all critical components for upgrades and maintenance. The two CPUs are mounted with solid passive heatsinks and the sockets are flanked on each side by 12 DIMM slots apiece.

Cooling is handled efficiently by a bank of 6 hot-plug fans behind the drive backplane and the entire assembly can be released and removed if required. Entry-level systems come with a single 3-slot PCI-Express riser card at the rear and there's room for a second 3-slot riser alongside, plus a 2-slot riser over the PSU bay.

On 2P systems, a large air shroud covers the CPU sockets and memory and is discarded to make way for the mezzanine board upgrade. This is a simple process, as the new board drops neatly into the metal support bracket and is secured by flipping down and locking its release handle.

HPE ProLiant DL560 Gen10: Storage choices

All models start with HPE's embedded Smart Array S100i controller which supports RAID 0, 1, 5, and 10 arrays for up to 14 SATA drives. It's not a great choice, as it only supports UEFI boot mode and Windows, doesn't cooperate with HPE's Intelligent Provisioning and won't talk to its integral Smart Array Administrator tool.

To get Windows Server 2016 on our system, we booted the server from a UEFI USB stick with the ISO and manually injected the Windows S100i driver during a standard installation. To avoid these issues, we recommend one of HPE's Smart Array snap-in modules which are far more amenable and offer advanced features such as SAS3 support, more RAID array options plus a choice of cache capacities.

Storage choices are extensive; the DL560 Gen10 employs HPE's standard 8+8+8 drive bay configuration layout, allowing up to 24 SFF drives to be installed in blocks. You can also fit HPE's universal media unit in the left bay which provides dual USB3, a DisplayPort plus optical drive and room for two SFF drives.

Up to 12 NVMe SSDs are supported with 2 in the first bay, 8 in the second bay and 2 more in the third bay. The mainboard also offers a pair of USB 2 ports and a microSD card slot for booting into an embedded hypervisor.

HPE ProLiant DL560 Gen10: Power, networks and management

Our system includes a single 1,600W hot-plug PSU and the bay has room for a second alongside. Dispense with the 2-slot riser and you can add another Flex Slot power backplane with two more PSUs connected to a separate power supply for extra redundancy.

A single locking tab on the right-hand PCI-Express riser means it can be released with a flick of the wrist and underneath you'll find HPE's FlexibleLOM slot. The price of our system includes a quad Gigabit module and HPE also offers a range of 10GbE and 25GbE adapters.

With HPE's iLO5 embedded management controller on the case, the DL560 Gen10 offers stiff platform security. This includes Secure Start, Secure Recovery and firmware validation using HPE's Silicon Root of Trust fingerprinting which will stop the server booting if it fails this test.

The iLO5 web interface is packed with useful information about critical components and if you upgrade to an Advanced or Essentials license, you'll get power metering, full OS remote control and virtual media services. We also added the server's iLO5 details to the lab's virtualized Hyper-V OneView app where we could closely monitor CPUs, power plus temperatures and use its tools for controlling power and running remote control sessions.

HPE ProLiant DL560 Gen10: Verdict

The ProLiant DL560 Gen10 is in a class of its own - it crams a remarkable Xeon Scalable package into a low-profile 2U chassis and teams this up with good storage choices and reassuringly tough platform security. It's ideal for businesses that want to pay only for what they need now but have the option to add more horsepower in the future as it can be upgraded from 2P to 4P capabilities in an instant.

Verdict

Growing businesses that want a low-profile Xeon Scalable server that can keep up with them will find a lot to like about HPE’s highly scalable ProLiant DL560 Gen10

Chassis: 2U rack

CPU: 2 x 14-core 2.2GHz Intel Xeon Gold 5120 (max 4)

Memory: 32GB ECC DDR4 2,400MHz (max 6TB, 2,666MHz)

Drive bays: 8 x hot-plug SFF (max 24)

Storage: No HDDs included as standard

RAID: HPE Smart Array S100i Gen10 SW RAID

Array support: RAID0, 1, 5, 10

Expansion: 3 x PCI-e Gen3 slots (max 8)

Network: 4 x Gigabit FlexibleLOM

Other: Internal USB 2 (2), microSD card slot

Power: 1 x 1600W hot-plug PSU (max 4)

Management: HPE iLO5 Standard with Gigabit

Warranty: 3yrs on-site NBD

Small and medium businesses (SMBs) looking for their first purpose-built 2U rack server or an upgrade for aging systems are seriously spoilt for choice. AMD's EPYC 9004 and Intel's Xeon Scalable fourth-generation processors present SMBs with a huge range of options allowing them to precisely tailor systems to their workloads.

The general consensus is that the 2U form factor offers SMBs the best combination of processing power, storage capacity and expandability. Even better, big name vendors such as Broadberry Data Systems, Dell and HPE all offer sizeable portfolios of 2U rack servers featuring a choice of AMD and Intel CPUs.

Along with a wide range of supported processors, these rack servers are an ideal solution for space-constrained SMBs. Offices are expensive to rent and 2U rack systems pack a remarkable amount of processing power into the smallest of spaces allowing businesses to save hard cash by consolidating services onto fewer servers.

How to choose a 2U rack server

Once you've decided to buy a rack server, the next question is which processor you should go for and it's important to pick the right one as these can represent the single largest outlay. There is an enormous range to choose from – combine all possibilities from AMD and Intel and you have over 70 products.

Another consideration is whether you need a single-socket (1P) or a dual-socket (2P) server. With the latest CPUs offering such high core densities, a 1P server may be able to deliver everything you need at a comparatively low cost and will be appealing to SMBs running VMware as it could reduce their per-processor licensing costs as one vSphere 8 licence supports up to 32 cores.

For sheer core density, AMD's EPYC 9004 CPUs are hard to beat with the family scaling from 32 cores up to a staggering 128 physical cores. Clearly, the mega-core EPYCs are overkill for most SMBs but if you do want AMD in the driving seat, consider the lower-cost 'P'-series models which are designed for 1P servers.

The Xeon Scalable CPU family currently comprises 49 models grouped into four 'metal' designations – Bronze, Silver, Gold, and Platinum. The ones of most interest to SMBs with light to moderate workloads are the Bronze, Silver, and possibly some of the more affordable Gold models.

The Silver CPUs are the sweet spot as they represent the best value with the Gen 4 range comprising three models offering 10, 12 and 20 cores and supporting DDR5 memory speeds of 4,000MT/sec. Smaller businesses on tight budgets may also want to consider Intel's family of affordable Xeon E-2300 CPUs as they are specifically designed for single-socket, entry-level servers with up to 128GB of DDR4 memory.

Thermal concerns

With great core count comes great power consumption, so when choosing your rack servers, check for chassis thermal restrictions in TDP as these may limit your processor choices. The Xeon Scalable Bronze 3408U has a 125W TDP, the three Silver CPUs range from 150W to 165W while the EPYC 9004 chips start at 200W and go up to a power-hungry 400W.

Most vendors have redesigned their server chassis to take into account the higher cooling demands of these fourth-generation CPUs. Good examples are Dell's PowerEdge R760xs which targets air-cooled environments and is restricted to Xeon Scalable Gold CPUs whereas the high-end PowerEdge R760 has a different internal design, can handle Platinum CPUs, and has options for liquid cooling.

Even the type of cooling fan needs to be considered with both Dell and HPE offering a choice of standard, high performance or very high performance fans. Check they use the same form factor as if you decide to upgrade the CPUs at a later date, you may need to replace the fan modules with high efficiency ones.

Storage decisions

Internal server storage can get eaten up quickly so it's essential you plan capacity for future demand to avoid costly upgrades and unnecessary downtime. This is where 2U rack servers score highly as they are available in a wide range of flexible configurations that can support large form factor (LFF) or small form factor (SFF) drives.

SATA drives offer the best price/capacity ratio with most 2U rack servers supporting at least twelve in their front bays with some also offering mid-plane and rear cages to increase the drive count even further. For greater storage performance, consider SAS HDDs or SSDs but these will be more expensive and may also require a SAS controller or suitable RAID card to support them.

Demanding workloads can take advantage of high-performance NVMe SSDs and you'll need to pay attention to the types of drive backplanes available. Some servers may require a dedicated NVMe SSD backplane whereas others offer tri-mode versions that support all types of drive interface.

Many blue chips charge a premium for their storage devices and may lock you in with strict warranty terms. However, vendors such as Broadberry Data Systems have no such conditions and expressly allow you to choose and fit your own storage devices without invalidating the warranty.

Serve and protect

Just as RAID provides protection against drive failures, it would be wise to specify dual power supplies so if one unit fails, the second will keep the server running. It's also advisable to invest in an uninterruptible power supply (UPS) to keep your data safe and allow the server to shut down cleanly if there's a power surge or outage. There are plenty of affordable models suitable for SMBs and APC (www.apc.com) has a useful selector tool on its website.

While RAID and power redundancy can protect you from hardware disasters, remember they aren't backup solutions. Get a proper data backup strategy running right from the outset – there are a wealth of on-premises and cloud solutions highly suited to SMBs and you can find plenty of in-depth reviews of all the main backup vendors here at IT Pro.

Remote management and monitoring are equally important as these allow support staff to track server health and keep them running reliably. The best rack servers have dedicated controllers on their motherboards which provide a web console so you can see the status of critical components in real-time and remotely control the server – Dell's iDRAC9 and HPE's iLO6 are standard setters in this area.

The latest 2U rack servers offer a stunning range of features including a choice of core-heavy AMD or Intel processing power, high-performance DDR5 memory, versatile storage arrangements, and a big expansion potential. Growing SMBs seeking their first rack server or a timely upgrade have never had it so good as they can choose the perfect system to suit current and future workloads and stay within their budgets.

FAQs

Dell EMC's PowerEdge R540 aims to offer growing businesses a fine blend of value and performance. This dual Xeon Scalable 2U rack server targets a wide range of apps including messaging, video surveillance, software defined storage or data backup and with support for the latest 12TB hard disks, has the storage capacity they demand.

The price of the review system may initially seem high but this includes a tasty pair of 2.1GHz Xeon Gold 6130 CPUs. Thermal limitations of the server's internal design restrict the R540 to 125W TDP CPUs but these 16-core brutes look capable of handling the most demanding apps.

The R540 comes with 8 or 12 LFF front drive bays which can be augmented with a two-bay carrier that slots in at the back. The server is geared up for storage space as although Dell EMC offers lower-capacity SFF drives and SSDs, these require an adapter to fit in an LFF carrier.

Base systems start with an embedded Dell PERC S140 chip supporting SATA devices and striped, mirrored or RAID5 arrays. The motherboard has a central riser for a RAID card and our system includes the H740P SAS3 model which adds RAID6 to the mix and comes with 4GB of battery protected cache memory (8GB will be enabled with a future firmware release).

A single hot-swap backplane handles all front drive bays and is cabled directly to the H740P controller. The rear drive cage is cabled to the backplane so it can be serviced by the same RAID card but it will reduce available PCI-Express slots to one.

You won't need any for network upgrades though, as the server supports a LOM (LAN on motherboard) card. These fit in a proprietary slot at the back and you can choose from dual Gigabit or 10GbE copper or SFP+ models.

The spare slot can be used for a BOSS (boot optimized storage solution) card which negates the need to put your OS on a hard disk. This PCI-Express card has dual M.2 SATA SSD slots, supports mirroring and costs 499 with two 240GB sticks.

Six cooling fans sit behind the backplane but are only the cold-swap variety. The R540 presents a busy interior and the two central fans are tricky to remove as their power connectors are obscured by the RAID card mounting bracket.

The server has an unusual memory arrangement as the first CPU gets ten DIMM slots while the second activates its six attendant slots. The R540 can use RDIMM or LRDIMM memory and with dual CPUs in residence, supports up to 512GB.

Remote management is excellent with the new iDRAC9 delivering a smart HTML5-based web interface and a wealth of operational information. Inventory, power consumption, CPU performance graphs -- it's all there and it offers tough platform security where a System Lockdown mode blocks any further configuration changes.

Data centre support staff will love the Quick Sync 2 module which is located in the server's left rack mount ear. We used it to link our iPad directly to the server over Bluetooth where the OpenManage Mobile (OMM) iOS app presented detailed views of server hardware plus options to reboot and remotely control it.

The OpenManage Essentials software provides general systems management but is now looking dated. Dell EMC's OpenManage Enterprise isn't due out yet but we fired up a Hyper-V VM of an early technical release and were impressed with the level of management features on offer.

The R540's internal design is a tad cluttered but it won't be faulted for features. The review system offers a mighty processing package, has a high storage potential and delivers classy server management tools.

Verdict

It's pricey but the R540 delivers a very powerful Xeon Scalable package with plenty of upgrade potential and a class management act

2U rack chassis;

2 x 2.1GHz Xeon Gold 6130

32GB 2,667MHz DDR4 (max 512GB)

Intel C622

Dell PERC H740P SAS3/4GB NVRAM cache/BBU

Supports RAID0, 1, 10, 5, 50, 6, 60

2 x 2TB Dell SATA LFF hot-swap drives (max 14)

3 x PCI-E 3.0 (max 6)

2 x Gigabit

1 x 1100W Platinum PSU (max 2)

Dell iDRAC9 Enterprise

3 years on-site NBD Basic warranty

Power 125W idle, 360W peak

Fujitsu's Server Primergy RX2540 M4 is ideal for SMBs that have future expansion and high growth on their minds. Along with support for every Xeon Scalable CPU model up to 205W TDP, this 2U rack server can be expanded to a massive 3TB of DDR4 memory, offers up to eight PCI-Express slots and has room for 28 SFF drives.

Our review system costs 4,390 and delivers a good specification to get you out the starting blocks. This includes a 10-core 2.2GHz Xeon Silver 4114 with room for a second CPU, 16GB of DDR4 memory and an 800W hot-plug PSU while its PRAID EP420i controller card brings SAS3 support and RAID5 and 6 arrays into the storage equation.

The server starts with an 8-bay SFF drive cage, to which two more can be added at the front and connected via SAS expanders or an extra RAID card. The full drive tally is realised by adding an optional 4-bay cage at the rear and Fujitsu is releasing a special order liquid-cooling kit allowing this particular specification to support the high-end 28-core Xeon Platinum 8180 CPU.

SMBs with tighter budgets can start small; an entry system with an 8-core 2.1GHz Xeon Silver 4110, 16GB of DDR4, an EP400i RAID card, eight SFF drive bays and a 450W PSU has an RRP of 2,216 ex VAT. The server also supports up to 12 LFF drive bays but the embedded Intel RAID chip only offers striped and mirrored arrays for eight SATA drives.

Cracking the lid reveals a beautifully designed interior with all components easily accessible for routine maintenance and upgrades. Cooling is handled efficiently by a bank of three hot-plug modules each containing two fans while the dual CPU sockets and 24 memory slots are covered by a solid transparent air shroud.

The RX2540 M4 offers a lot of expansion options with six embedded PCI-Express slots that can be increased to eight using optional riser modules. These will also be required if you want to fit up to two Nvidia Quadro or Tesla GPGPU cards which reduce CPU support to models with a maximum 150W TDP.

The base server comes with dual Gigabit network ports and the price we've shown includes Fujitsu's quad-Gigabit DynamicLOM adapter card. This snaps into a proprietary slot on the motherboard, and other models are available with dual Gigabit, dual copper 10GbE and dual or quad fibre 10GbE ports.

Power choices are plentiful as along with the 800W Platinum PSU in our server, you have a spare bay for a second redundant supply and Fujitsu also offers 450W, 1200W and 1300W Platinum or Titanium variants. Overall power consumption for our system was good with it drawing 70W in idle and peaking at only 135W with the Xeon Silver under a 100 percent load.

Fujitsu's ServerView systems management software suite is still looking in need of a major update but the server shows off Fujitsu's new iRMC S5 controller. Its freshly-designed web console offers tighter access security and makes all critical system information much more accessible than its predecessor.

It provides a complete breakdown of power consumption for key components plus detailed event logs and alerting facilities. The price we've shown includes the iRMC advanced license which activates video redirection for full OS remote control, virtual media services and video capture.

The Primergy RX2540 M4 is an ideal choice for fast growing SMEs that want a Xeon Scalable server that can handle all their current and future demands. Remote management features are average, but this rack server offers a good hardware package for the price and a truly remarkable expansion potential.

Verdict

The Fujitsu Server Primergy RX2540 M4 is an ideal choice for fast growing SMEs that want a Xeon Scalable server that can handle all their current and future demands. Remote management features are average, but this rack server offers a good hardware package for the price and a truly remarkable expansion potential.

2U rack chassis

1 x 2.2GHz Intel Xeon Silver 4114 (max 2)

16GB DDR4 RDIMM (max 3TB)

Intel C624

Fujitsu PRAID CP420i SAS3/2GB cache/BBU

Supports RAID0, 1, 1E, 10, 5, 50, 6, 60

3 x 300GB SAS3 SFF hard disks (max 28)

2 x Gigabit; 4 x Gigabit DynamicLOM card

6 x PCI-E 3.0 (max 8 with risers)

1 x 800W hot-plug PSU (max 2)

Fujitsu iRMC S5 Advanced with Gigabit

3 years on-site NBD warranty

Power 70W idle, 135W peak

Lenovo revealed its intent on the server market last year with one of the biggest product launches in its history. The ThinkSystem family introduces twelve new servers with the SR550 on review representing Lenovo's opening gambit on the cost-conscious SME sector.

Along with a completely revamped hardware design and support for Xeon Scalable CPUs, Lenovo has radically improved its server management offerings. You'll see this on first power up as the SR550 presented us with the new XClarity Controller (XCC) interface on its dedicated management port.

The XClarity Provisioning Manager handles OS deployment with aplomb. We selected its assisted installation routine during boot up and it helped get a RAID array created and Windows Server 2016 loaded inside 30 minutes.

The XCC remote web interface presents a slick home page offering a wealth of information about system health and critical components. It provides quick access to remote control and presents coloured graphs showing utilization and system wide power usage.

Along with hardware inventory, it provides a complete breakdown of component utilization, fan and voltage status plus temperatures. RAID controllers can be directly configured from the XCC interface which also provides firmware update tools and options to set access security.

We run Lenovo's XClarity Administrator 1.4 in our lab as a Hyper-V host and after adding the SR550 to its console, we could remotely monitor it, use its email alerting services and employ its image library for firmware updates and OS deployments. There's more: we used Lenovo's free XClarity Mobile iOS app on our iPad to remotely access the Administrator host and monitor our server on the move.

The SR550 is solidly constructed and offers plenty of storage options with support for up to 12 LFF or 16 SFF drives. Our review system was supplied with 8 LFF drive bays plus the operator panel and the price includes two 1TB 'simple-swap' (cold-swap) SATA hard disks.

RAID options start with the embedded Intel RSTe controller which supports up to 8 SATA drives plus striped, mirrored and RAID5 arrays. Lenovo offers an optional RAID 530-8i adapter which introduces SAS3 support, while the 930-8i and 930-16i add RAID6 and 2GB of cache memory.

The review system includes an entry-level 6-core 1.7GHz Xeon Bronze 3104 and has a second CPU socket ready for future upgrades. Thermal design limits the SR550 to CPUs with a maximum 125W TDP and 20 physical cores.

This is unlikely to trouble its target market, as it supports all Bronze and Silver models plus a wide range of Gold CPUs. The price includes 16GB of TruDDR4 memory and with both CPU sockets occupied, you can use the 12 DIMM slots to expand this to 384GB of RDIMM or 768GB of LRDIMM memory.

Our system had a single 750W Platinum PSU which can be augmented with a second for hot-plug redundancy and Lenovo also offers 550W Platinum and 750W Titanium versions. The low 85W TDP of the Bronze CPU was reflected in our power tests with the server drawing 57W in idle and peaking at a mere 70W under extreme load.

Only a single PCI-Express slot is available in base systems but adding two optional risers increases the slot count to six. You get dual Gigabit plus a dedicated XCC Gigabit port and the server supports Lenovo's LOM adapters with a choice of dual Gigabit or dual 10GbE copper or SFP+ fibre ports.

The ThinkSystem SR550 offers SMBs an affordable entry point into the world of Xeon Scalable processing. Its flexible design allows you to start small and easily expand to cope with increased workloads while Lenovo's new server management solutions won't be bettered.

Verdict

The ThinkSystem SR550 offers SMBs an affordable entry point into the world of Xeon Scalable processing. Its flexible design allows you to start small and easily expand to cope with increased workloads while Lenovo's new server management solutions won't be bettered.

2U rack chassis;

1.7GHz Xeon Bronze 3104 (max 2);

16GB 2,667MHz TruDDR4 (max 768GB with LRDIMMs);

Intel C622;

Intel RSTe;

Supports RAID0, 1, 5, 10;

2 x 1TB SATA simple-swap drives (max 12 LFF/16 SFF);

1 x PCI-E 3.0 (max 6);

2 x Gigabit;

XCC Enterprise with 1GbE;

750W hot-plug PSU (max 2);

3yrs on-site NBD warranty;

Power 57W idle, 70W peak

Lenovo is upping its game in the storage market as its latest ThinkSystem DS series of arrays aim to offer an enterprise-class feature set and performance at a price mid-sized businesses can afford. On review we have the top-dog DS6200 a 2U appliance that targets I/O intensive applications and offers end-to-end 12Gbits/sec SAS support, real-time data tiering, high availability and a mighty expansion potential.

Along with 24 SFF drive bays, the DS6200 is endowed with dual controllers running in active/active mode. Each as a generous 16GB of cache memory, perform low-latency cache mirroring and team up a super-capacitor and Flash memory to protect their contents during a power outage.

Our review system has Lenovo's CNCs (converged network controllers) installed, which each have four SFP+ ports and support both 8/16Gbits/sec Fibre Channel and 1/10GbE iSCSI operations. You need to manually set their personality using the CLI but you can have them as all FC, all iSCSI or a mixture of both where the first pair are FC and the second pair are iSCSI.

High-speed SAS3 support extends to the CNC's expansion ports. You can daisy-chain up to nine disk shelves from them using fault-tolerant links for a total of 240 SFF or 108 LFF drives, and mix and match different shelves as well.

Lenovo ThinkSystem DS6200 review: Installation

Deployment doesn't get any easier as the appliance's web console runs a wizard that handles enabling management access and configuring system settings including SNMP monitoring. Out of the box, our CNCs had their port personalities set to iSCSI but we wanted FC support for our performance tests.

It's easy enough to change and we used the PuTTY client to SSH to the appliance and from its CLI, used the set host-port-mode' command to change them all to FC. Next up is virtual storage provisioning, where the appliance manages two storage pools with one assigned to each controller.

We were supplied with eight 800GB SSDs for testing and quickly created a four-drive RAID10 pool, or disk group, for each controller. As we were only using SSDs, the appliance placed them in the Performance tier and if you add SAS hard disks, they'll go in the Standard tier while Midline SAS drives are automatically allocated to the Archive tier.

The base license only supports real-time data tiering across SAS and ML-SAS HDDs and to use all three tiers, you'll need an optional license which costs around 2,553. However, the base license does enable SSD read caches which can be assigned to disk groups and are not part of the tiering process.

Lenovo ThinkSystem DS6200 review: FC performance

For performance testing, we used two Xeon Scalable and two Xeon E5-2600 v4 Windows Server 2016 hosts, all equipped with QLogic single-port 16Gbits/sec FC adapters. We mapped a dedicated virtual volume to each server and connected two servers to Controller A and the other two to Controller B.

Our first test using a single server delivered excellent results with Iometer reporting steady sequential read and write rates both of 12.3Gbits/sec. Random read rates held steady while write speeds dropped to 7.5Gbits/sec.

With all four servers in the Iometer mix, we recorded top cumulative sequential reads and writes both of 49Gbits/sec. Total random read and write speeds settled at 48Gbits/sec and 30Gbits/sec with each server recording average I/O response times of less than 0.3ms

I/O throughput is excellent as Iometer recorded cumulative sequential read and write rates of 388,000 and 115,000 IOPS. Moving to random operations saw read and write throughput drop to 228,000 and 114,000 IOPS with latency on all four servers hovering around the 1.2ms mark.

Lenovo ThinkSystem DS6200 review: Snapshots and management

The base license includes snapshot support for up to 128 targets which can be optionally increased to 512 or 1024. They're easy to manage and can be run on-demand or scheduled as often as every minute within a set daily time window, with an option to retain up to 32 for each volume.

We found snapshot recovery is equally swift as we selected a volume and viewed its snapshots in the list below. After choosing the one we wanted, we disconnected any hosts, selected the Rollback option and had our deleted data returned to its rightful place in a matter of seconds.

Asynchronous replication is another optional feature and a bonus is the DS6200 can replicate volumes to any remote ThinkSystem DS or older Storage S series appliance. It requires iSCSI connections to function and can be set to run immediately and as often as every hour.

We found the appliance's web interface very easy to use, and it provided quick access to the various storage functions. It also provides a handy performance section where we could monitor all components ranging from a single drive or host port to an entire controller.

We run Lenovo's XClarity Administrator as a Hyper-V VM in the lab and it automatically discovered the DS6200. Once we'd supplied our management credentials, it was amalgamated into the Storage section of the XClarity web console where we could view a full hardware inventory, monitor power consumption and temperatures, remotely control power and watch out for system alerts.

Lenovo ThinkSystem DS6200 review: Verdict

Mid-sized businesses that baulk at the cost of all-Flash storage arrays can rest easy with Lenovo's ThinkSystem DS6200. It offers a host of affordable enterprise storage virtualization features and teams them up with integral real-time data tiering, great Flash performance and plenty of room to expand capacity.

Verdict

The ThinkSystem DS6200 is a fine choice for mid-sized businesses that want the best storage performance, real-time data tiering and scalability at a sensible price

Chassis: 2U rack

Storage: 24 x SFF hot-swap drive bays

Power: 2 x 580W hot-plug PSUs

Two controllers each with the following:

Memory: 16GB cache with super-capacitor and Flash memory

RAID support: RAID1, 10, 5, 6

Data ports: 4 x SFP+ supporting 8/16Gbps FC and/or 1/10GbE iSCSI

Management: Gigabit Ethernet

Expansion: 12Gbps SAS3 ports (max 108 LFF/240 SFF drives)

Other ports: Mini-USB CLI port

Snapshots: Base 128 (optional 512 or 1024)

Management: Web browser

Warranty: 3 years on-site 9x5 NBD

Options: SSD Data Tiering License (01GV561), £2,553 ex VAT

The Xeon Scalable family is Intel's biggest product launch for nearly eight years and it augments this huge range of server CPUs with a feast of motherboards and rack platforms. In this exclusive review, we look at the CyberServe Xeon SP2-R2224 from Broadberry Data Systems - an all-Intel system aimed at SMBs and enterprises with power, storage and growth on their minds.

The CyberServe is based on an Intel Server System R2224WFTZS 2U rack platform sporting an S2600WFT motherboard. This partnership offers an impressive range of features as along with support for the entire Xeon Scalable family, it has room for 24 SFF hot-swap drives at the front and two more at the rear using an optional cage.

It's a powerful package, and the price includes a tasty pair of 2.2GHz Xeon Gold 5120 CPUs offering 14 physical cores apiece. These are teamed up with 64GB of 2,400MHz DDR4 memory, which can be boosted to 1.5TB.

The embedded C624 chip supports up to twelve SATA drives, two M.2 SATA SSDs and RAID0, 1 and 5 arrays, but the CyberServe also has Intel's RMS3CC040 RAID controller nestling in its dedicated slot. This adds SAS3 plus RAID50, 6 and 60 to the mix, along with 1GB of embedded cache.

The system includes two 240GB Intel SATA SSDs and all 24 front bays are available, as the RAID card is connected to a SAS expander behind the drive backplane. The backplane also features 24 OCuLink ports for direct-attach NVMe SSDs.

The motherboard has four embedded OCuLink ports which can be connected to the backplane using Intel's optional SFF-8611 cable kits. You can have more than four NVMe drives, as Intel offers PCI-Express Switch AIC adapters with four or eight OCuLink ports.

Network options get off to a flying start as the motherboard has two embedded 10GBase-T ports. There are eight free PCI-Express slots but you won't need them to expand networking, as Intel's OCP expansion modules fit in a dedicated slot at the rear and offer dual 10-Gigabit copper or SFP+ ports.

Internal cooling is handled well; the CPUs are covered with a solid transparent shroud and looked after by six hot-plug cooling fans in front of the motherboard. It needs to be good, as the system is validated for CPUs with 205W TDPs so it can handle the top-dog 28-core Platinum 8180s.

The CyberServe has a big power envelope to match - it comes with dual 1.3kW hot-plug power supplies. Even though the Gold 5120 CPUs have a 105W TDP, we found power consumption to be reasonably low with the system drawing 110W in idle and 286W under extreme load.

Remote server management is present as the CyberServe has Intel's BMC controller, which activates a dedicated network port at the rear. Its web console can't match HPE's iLO5 or Dell's iDRAC9 for features but it does provide sensor views of critical components and issues email alerts if values exceed their predefined thresholds.

The BMC also maintains an event log, has a new view added for NVMe drives and includes full remote control as a standard feature and not as an option. Installed locally, Intel's Active System Console (ASC) also provides plenty of valuable information about hardware while the server is running.

ASC is a web service so you can access it remotely via a browser, keep an eye on the hardware and link its email alerting service with multiple policies. Last up ts Intel's freshly-designed RAID Console 3 which also provides local and remote web browser access to the server's storage controllers, arrays and drives.

The CyberServe Xeon SP2-R2224 is an impressive 2U rack server and its versatility makes it well suited to a wide range of data centre duties. Combining this with its innovative storage features and a low price the blue chips won't come close to matching puts Broadberry's powerhouse firmly on the IT Pro A-List.

Verdict

The CyberServe Xeon SP2-R2224 is an impressive 2U rack server and its versatility makes it well suited to a wide range of data centre duties. Combining this with its innovative storage features and a low price the blue chips won’t come close to matching puts Broadberry’s powerhouse firmly on the IT Pro A-List.

2U rack chassis

Intel Server Board S2600WFT

Intel C624 chipset

2 x 2.2GHz Intel Xeon Scalable Gold 5120 CPUs

64GB 2,400MHz DDR4 (max 1.5TB)

Intel Integrated RAID RMS3CC040/1GB DDR3 cache

supports RAID0, 1, 10, 5, 50, 6, 60

2 x 240GB Intel S3520 Enterprise SSDs (max 26 SFF)

2 x M.2 SATA

2 x 10GBase-T

Intel BMC with 10/100

8 x PCIe x3 slots

Intel OCP slot

2 x 1.3kW hot-plug PSUs

3yrs on-site NBD warranty

Power 110W idle, 286W peak

Modular servers offer SMBs and enterprises an affordable and flexible means of matching infrastructure with workload. The SD530 is Lenovo's first modular server in the ThinkSystem family and is built from the ground up to support the latest Xeon Scalable CPUs.

This rack dense solution comprises Lenovo's 2U high D2 enclosure with room for up to four SD530 server nodes. Each node is designed to be completely independent and has two CPU sockets, up to 16 DIMM slots, dual embedded 10GbE ports, an integral RAID controller and up to six hot-swap SFF drive bays.

Virtually everything you'd expect to see on Lenovo's standard ThinkSystem servers is present in the SD530 nodes. They can handle SATA and SAS drives plus NMVe SSDs, will accept a dual M.2 SSD enablement kit and sport Lenovo's XCC (XClarity Controller) for slick remote management and monitoring.

Chassis and configuration

The D2 enclosure handles all power and cooling and has a passive mid-plane which the nodes connect to when inserted. Our review system included two big 2,000W Platinum hot-plug PSUs and Lenovo also offers 1,600W and 1,100W options.

Cooling is handled by a central bank comprising three hot-plug dual-rotors, flanked on each side by cold-swap modules each with two large fans. The whole assembly can be accessed easily by removing the hatch at the rear of the lid.

Called an I/O shuttle, this self-contained assembly contains both power supplies, up to eight PCI-Express slots, the chassis SMM (system management module) and an 8-port EIOM (Ethernet I/O module). Each node is assigned PCI-Express slots in fixed locations on the risers and in all cases, the I/O shuttle must be removed to install cards with the two inner slots on the x8 shuttle requiring the central switch module to be removed as well.

Node design

Lenovo offers a wide range of warm-swap node models and the total memory and drives supported will be determined by your choice of CPU. The single node in our review system is equipped with dual 2.6GHz 12-core Xeon Gold 6126 CPUs which require larger heatsinks, thus reducing the number of DIMM slots to twelve and maximum 2,667MHz TRuDDR4 memory to 768GB.

Ranging from the entry-level Bronze 3104 to the high-end Platinum 8153, twelve CPU models can use smaller heatsinks and these nodes have 16 DIMM slots allowing memory to be pushed to 1TB. Cooling requirements also affect the number of available drive bays so for maximum storage capacity, choose cool-running CPUs that support the enclosure's 2x3 backplane.

The node motherboard has an embedded Intel RSTe controller supporting six cold-swap SATA drives. For 12Gbps SAS drives, you have the 430-8i which has no RAID or the 530-8i which supports mirrors, stripes and RAID5, self-encrypting drives and FastPath acceleration.

SMM configuration

The enclosure supports two types of SMM where one has a single network port for managing all four nodes. The other type has dual network ports allowing multiple enclosures to be daisy-chained and all managed via a single IP address.

The SMM not only provides its own web management interface but also facilitates remote access to each node's XCC interface. By default, it allows access to the node's XCC interfaces but its own web server is disabled and proved to be tricky to get working.

To enable it, you need to use the IPMITool or a similar utility for issuing direct IPMI commands. This is made more complicated as the 'enable' parameter appended to the command is in hex (which is in the manual) and has to be issued to a node's XCC IP address.

Suffice to say it took us a while to get it working but we could then access the SMM web console. This is fairly basic but does provide plenty of information about installed components, power and cooling plus direct access to each node's XCC interface and options for issuing email alerts.

Node management

For local node management, you can specify the optional KVM module which occupies one drive bay. Remote access will probably be enough though, as the node's XCC controller delivers a wealth of remote management and monitoring features.

The XCC home page provides details on all critical components and firmware versions, access to remote control and a heap of graphs revealing utilization and power usage. It provided us with direct access to the M.2 SSDs and the RAID controller for storage configuration, a complete inventory of all hardware and facilities to define server, firmware and access security.

We had no problems managing it with the lab's XClarity Administrator 1.4 Hyper-V VM which auto-discovered the node's XCC. We could monitor power usage for the entire node, CPUs or memory, employ the image library to push an OS to it and use Lenovo's xClarity Mobile iOS app on our iPad to remotely access the Administrator host and monitor the node on the move.

Verdict

The review server with one node may look pricey, but bear in mind the dual Gold 6120 CPUs in our node represent over 60% of the price for the total system. Support for nearly all the Xeon Scalable family means it's easy to tailor the hardware to meet your budget, making the ThinkSystem SD530 an affordable and versatile modular server solution for enterprises looking to increase processing density in their data centre.

Verdict

The ThinkSystem SD530 modular server is ideal for data centres that want an affordable rack dense solution with the latest Xeon Scalable CPUs

(As reviewed)

Chassis: Lenovo D2 2U rack

Expansion: 4 x compute node bays

Power: 2 x 2000W hot-plug PSUs

Network: 8 x 10GBase-T module (2 ports per node)

Cooling: 8 x hot-plug fan modules

Expansion: 8 x PCI-Express 3 slots (2 per node)

Management: Lenovo SMM with Gigabit/USB

One hot-swap SD530 node with the following:

CPU: 2 x 2.6GHz 12-core Xeon Gold 6126

Memory: 64GB TruDDR4 (max 768GB)

Storage: 2 x internal 32GB M.2 SSDs, 2 x 240GB Intel SATA SFF SSDs (max 6)

RAID: Lenovo ThinkSystem RAID 530-8i Dense SAS3

Array support: RAID0, 1, 10, 5, 50

Network: 2 x 10GbE

Management: XCC Enterprise, XClarity Administrator

Warranty: 3yrs on-site NBD

2017 has been an excellent year for technology and hardware. Consumers have been wowed by the likes of the Samsung Galaxy S8 and a list of redesigned Apple products, while businesses have been able to take advantage of unparalleled processing power thanks to servers from companies like HPE and Dell.

With such a wealth to choose from, it's hard to pick which products have been most impressive, but with some careful thought, we've sifted through the year's launches to bring you IT Pro's top products of the year.

Best laptop

Winner - Apple MacBook Pro

Apple's top-end MacBook models are widely regarded as the gold standard in business-class notebooks, and with good reason. Combining a sleek and stylish design with nigh-unbeatable performance thanks to 2017's upgraded Kaby Lake model, the MacBook Pro is an absolute powerhouse for everyone from developers to designers.

It's not entirely without problems - we wish Apple had included a couple of ports besides Thunderbolt 3, for example, and the Touch Bar version has some battery life problems - but features like the ultra-shallow keyboard and mammoth trackpad set it apart from its competition.

Highly commended - Lenovo ThinkPad X1 Carbon

Another heavyweight in the enterprise hardware space, Lenovo's venerable ThinkPad range has a sterling pedigree. The ThinkPad X1 Carbon is sturdy, reliable and well-suited to tasks of all stripes, with a truly excellent keyboard and finely-tuned features for businesses.

Best tablet

Winner - Apple iPad Pro

Apple has thoroughly dominated the tablet computing category this year - and not just because it's been one of the only companies to release a premium tablet. The company's devices have been sophisticated, powerful and portable, and the iPad Pro is the perfect combination of all three.

The Apple Pencil support made it an excellent choice for designers and artists, but now that Apple has introduced a file manager and a suite of proper multi-tasking capabilities, the iPad Pro can genuinely be considered as a laptop replacement. Showcasing the very best of Apple's design abilities, both the 10.5in and 12.9in iPad Pro models are the pinnacle of tablet design.

Highly commended - Apple iPad

If you can't quite bear to part with the 620 starting price of the iPad Pro, Apple's entry-level iPad makes a handy compromise. While it doesn't have the full range of productivity features seen on the Pro range, the basic iPad is still sleek and speedy enough for day-to-day use.

Best 2-in-1

Winner - Microsoft Surface Pro

One look at how many companies have aped the design of Microsoft's flagship convertible should give you a good indication of how popular the Surface Pro is. The kickstand design is elegant and intuitive, and the keyboard is one of the most satisfying typing experiences we've had, whilst also being amazingly thin and light.

There's no compromise on components, either; the 12.3in screen is pin-sharp and absolutely gorgeous to boot, and one of Intel's seventh-generation processors provides a serious amount of oomph. Forget compromise; this 2-in-1 does everything a laptop can, and then some.

Highly commended - HP Elite X2 Folio

While it's definitely borrowing some design inspiration from the Surface Pro, the HP Elite X2 Folio brings a few things to the table that Microsoft's device doesn't. It's user-upgradeable, for example, which could end up saving corporate IT departments a whole heap of time and money in the long term.

Best Chromebook

Winner - Google Pixelbook

Chromebooks aren't just cheap and cheerful laptops for schools, and Google's out to prove it with the Pixelbook. A razor-thin body and eye-catching design is backed up by surprisingly powerful internal hardware and all the AI-powered software features you'd expect from a flagship Google product.

What's more, improvements to Chrome OS and the ability to run Android apps mean that the Pixelbook is (very nearly) as versatile and capable as a standard Windows laptop. It's expensive, but it's liable to inspire serious notebook envy wherever you go.

Highly commended - HP Chromebook 13

At around 600, the HP Chromebook 13 is much more affordable than the pricey Pixelbook, without making too much of a compromise on design or performance. Like most Chromebooks, it'll struggle with seriously heavy-duty workloads, but for general tasks, this lightweight machine should perform very well indeed.

Best smartphone

Winner - Samsung Galaxy S8

2017 has been a truly excellent year for smartphones, and the Samsung Galaxy S8 has been firmly leading the charge. The first major smartphone out of the gate with an edge-to-edge OLED display, the S8 set an example followed by companies like OnePlus, Honor and even Apple.

Combine this with one of the best smartphone cameras we've ever seen, super-quick internals and a very capable battery life, and it all adds up to a device that's very attractive indeed. Move over, iPhone; there's a new king in town.

Highly commended - Google Pixel 2

The follow up to last year's surprise hit, the Google Pixel 2 is a deeply impressive piece of technology powered by intelligent AI features and smart UI design. Its real selling-point is an absolutely world-class camera, which has yet to be beaten by any other smartphone. Were it not for a slightly underwhelming screen, the Pixel 2 would be topping our list this year.

Best 1U server

Winner - Dell EMC PowerEdge R640

The Dell EMC PowerEdge R640 is part of the company's recently-launched 14th generation of servers, and Dell has pulled out all the stops to cram this space-efficient 1U rack server full of storage and power. Not only does it have full support for Intel's Xeon Scalable processors, it's also got some mightily impressive storage options.

On top of that, Dell's iDRAC9 management console is a smash hit, offering a comprehensive overview of operational information, security-centric features and outstanding remote management and administration capabilities. It's every bit the equal of HPE's superb iLO5 controller.

Highly commended - Lenovo ThinkSystem SR630

Expansion potential is the name of the game here, with Lenovo's ThinkSystem SR630 offering plenty of internal space for further upgrades, despite a compact 1U form factor, and Xeon Scalable processor support for improved performance. More importantly, the XClarity Controller will make remote management a breeze.

Best 2U server

Winner - Dell EMC PowerEdge R740xd

The first server in Dell EMC's 14th-generation PowerEdge line, the PowerEdge R740xd is an excellent example of what customers can expect from the company now that the merger between Dell and EMC has been completed. Remote management capabilities have been improved, and EMC's expertise has been put to use with a superb array of storage features.

It's also a powerful showcase of Intel's relatively new enterprise-grade Skylake-SP Xeon processors. The R740xd supports the whole family, all the way up to Intel's 28-core Platinum Xeon monster, and its sheer processing power is tough to beat. Blue-chip vendors, take note: this is how it's done.

Highly commended - HPE ProLiant DL380 Gen10

Like the PowerEdge, HPE's tenth-generation flagship 2U server features Xeon Scalable processor support and extensive storage options. It's an improvement over the previous iteration in just about every way, and it's got the huge room for expansion too, making it a great option for growing businesses.

Best tower server

Winner - HPE ProLiant ML110 Gen10

Smaller businesses on a tighter budget may think that cutting-edge Xeon Scalable servers are out of their reach, but not anymore. The 10th-generation update to HPE's ProLiant ML110 tower server brings SMBs excellent performance coupled with generous expansion options.

It also comes with HPE's outstanding iLO5 management controller, providing plenty of remote management, monitoring, and security features. For smaller organisations that want a heavyweight server without the top-end price tag, this is an excellent choice.

Highly commended - Fujitsu Server Primergy TX1320 M3

If you're tight on space, Fujitsu's smallest-ever tower server will be right up your street. Not only that, but it's also great value. Combine that with beefy Xeon processors and heaps of clever design features to make the most of the limited internal space, and you've got a very capable package.

Best security appliance

Winner - WatchGuard Firebox T70

The biggest thing that the WatchGuard Firebox T70 has going for it is its virtually unbeatable price. Not only is it extremely competitive, it's also one of the highest-performing desktop appliances we've ever seen, with 4GBits/sec firewall throughput.

It's bundled with a year's subscription to WatchGuard's enterprise-class software suite too, including anti-spam, anti-virus and web-content filtering, as well as a Gold-level support subscription. Lastly, it has top-notch monitoring features for ultimate peace of mind.

Highly commended - Kerio Control NG300W

The Kerio Control NG300W is one sturdy UTM appliance, with four Gigabit Ethernet ports, a 2.4GHz Intel Atom CPU, and an inbuilt heatsink. It doesn't fall down on security features either, offering a comprehensive and well-rounded package. The only absence is anti-spam features, but otherwise, it's an excellent security appliance.

Best storage array

Winner - HPE MSA 2052 Storage

HPE has raised the bar with its 5th-gen MSA 2052 storage appliance, doubling the performance of an already-excellent array whilst keeping it at the same wallet-friendly price as its predecessor.

Performance is accordingly excellent, and it's twinned with a load of powerful features to make virtualisation, deployment, and recovery as easy as possible. Considering that it includes HPE's full suite of data tiering capabilities as well, it's a steal.

Highly commended - Broadberry CyberStore 224S-WSS rack

Despite going up against titans like HPE, Dell EMC, and Fujitsu, Broadberry was first out of the gate with a Windows Storage Server 2016 appliance - and for a very attractive price, too. On top of that, the Broadberry CyberServe 224S-WSS comes with 26 hot-swap SFF drive bays, heaps of network connection options, twin Intel Xeon E5-2620 v4 processors and 64GB of RAM.

Best NAS drive

Winner - Qnap TS-1685 desktop

For those that want high-volume desktop storage capabilities, it's hard to beat Qnap's TS-1685. Not only does it feature 16 hot-swap drive bays, it's also got blazing network performance thanks to its four 10GbE Ethernet ports, and support for expansion via additional PCIe cards.

You'll be hard-pressed to match its performance in terms of disk read and write speeds as well. Add in a whole heap of backup and storage features, and the TS-1685 is a superb addition to any small office setup.

Highly commended - Netgear ReadyNAS RN426

Data protection is top of the priority list for this SMB-friendly NAS drive. Backup performance is excellent, with cloud sharing, disaster recovery and snapshot features aplenty. It's also nice and hardy, thanks to its rock-solid steel construction.

Best printer

Winner - Xerox VersaLink C400DN

The first thing that's likely to attract SMBs Xerox's VersaLink C400DN is the low price, but one of the most useful features is the large variety of onboard apps available for it, including heaps of cloud and network printing tools. It's also one of the rare printers that supports wired and wireless networking simultaneously.

Print quality is absolutely superb, delivering razor-sharp and detail-packed pictures and documents at exceptional speeds. As if that wasn't enough Xerox offers a lifetime on-site warranty - all you have to do is buy a toner pack in the last 60 days of your warranty period, and the company will renew it for another year. Can't say fairer than that.

Highly commended - Brother MFC-L8900CDW MFP

Although its initial purchase price isn't as attractive as some of its rivals, this A4 laser printer makes up for it with a battery of excellent cloud and security features, combined with speedy, high-quality prints. Running costs are reasonable, too, so TCO shouldn't be a problem later down the line.

In its broadest product launch since acquiring IBM's System x portfolio, Lenovo has gone back to the drawing board and completely redesigned its ThinkSystem servers from the ground up. They deliver a wealth of new features and in this exclusive review we bring you a first look at its ThinkSystem SR650 general purpose 2U rack server.

Support for Intel's Xeon Scalable CPU family is at the top of Lenovo's agenda and the SR650 shows off a new storage design offering more choices. Along with support for up to 14 LFF or 24 SFF drives, the SR650 features embedded NMVe PCI-Express ports as standard, while the AnyBay carriers from previous generations increase storage versatility.

A welcome update for us is Lenovo's embedded XCC (xClarity Controller) chip. Remote server management has been a major weakness of current System x and ThinkServer platforms and the XCC puts it firmly alongside Dell's iDRAC9 and HPE's iLO5 for features.

Management

The first benefit of the ThinkSystem firmware became apparent when we powered the SR650 up; boot times have been drastically reduced. Our lab's System x3550 M5 rack server takes a yawn-inducing 6 minutes to go from power on to OS login, whereas the SR650 cut our waiting time to 110 seconds.

OS deployment is easier too, as the xClarity Provisioning Manager provides an assisted installation routine. Selected during boot up, it helped us get a RAID array created and Windows Server 2016 loaded with all the correct drivers inside 30 minutes.

There's much more, as the XCC web interface is like nothing we've seen from Lenovo before. It's slick home page opens with a wealth of information about all critical components, quick access to remote control, a rundown on all firmware versions and a set of coloured blocks showing utilization and system wide power usage.

It provides a full inventory of every hardware component and a detailed breakdown of power usage, component utilization, fan and voltage status plus system temperatures. You also have direct configuration access to the RAID controllers, firmware update tools and options to set server, firmware and access security.

After updating the lab's xClarity Administrator Hyper-V host to v1.3.2, we were able to add the SR650 to its console, remotely monitor it and use its image library for centralized OS deployment. We also used Lenovo's xClarity Mobile iOS app on our iPad to remotely access the Administrator host and a new feature for cellular-equipped mobiles is the ability to tether them directly to the ThinkSystem's front USB port for local monitoring.

Storage

The server's front panel uses the popular 8+8+8 drive bay configuration, allowing it to support from 8 to 24 SFF drives (HPE's DL380 Gen10 supports 30 SSF drives). The SR650 LFF model has room for up to 12 drives at the front and this chassis also supports an optional two-bay cage fitted at the rear in place of the 3-slot PCI-Express riser.

Lenovo's AnyBay is unique as it supports SAS, SATA, standard SSD or NVMe PCIe SSD interfaces in the same bay. With the appropriate AnyBay backplane factory fitted (and dual CPUs), the SR650 16-bay model can handle up to eight of these bays.

The clincher is the pair of embedded dual-port NVMe PCIe connectors on the motherboard. Other server vendors require you to purchase an optional adapter card for NVMe drives but the SR650 supports four out of the box at no extra cost -- the extra four require Lenovo's optional NVMe switch adapter.

RAID choices are plentiful with the review system including a ThinkSystem 930-8i SAS3 adapter fitted in the dedicated PCI-Express slot in the centre of the motherboard and including 2GB of battery protected Flash cache. Lenovo also offers a dual M.2 SATA SSD enablement kit and two mirrored 128GB modules for only 328.

Expansion and power

Internally, the server is well designed with easy access afforded to all major components. The SR650 supports the entire Xeon Scalable family with the price of our review system including dual 8-core 2.1GHz Xeon Silver 4110 CPUs partnered by 32GB of TRuDDR4 memory.

The Xeon Silver CPUs are aimed at entry-level data centre applications with all five models supporting 2,400MHz memory, 768GB per socket plus Intel's HT and Turbo Boost technologies. Their low 85W TDP was reflected in our power tests, and the server drew 80W in idle and a mere 118W under extreme load.

The server was supplied with the full complement of six internal hot-swap cooling fans and dual 750W Platinum hot-plug PSUs. Other PSU options include 550W, 750W Titanium, 1100W and 1600W and all share a common form factor across the entire ThinkSystem range.

Instead of embedded network ports, the motherboard provides a LOM slot which links up with the C624 PCH and supports dual or quad Gigabit and 10GbE cards. Lenovo's Infiniband and 25/40/100GbE cards require a separate riser with an ML2 slot.

Conclusion

The ThinkSystem SR650 represent a huge step forward for Lenovo as it's well designed, good value and highly expandable. Support for the Xeon Scalable family allows it to be customized to suit any workload, storage features see big improvement and the new xClarity Controller finally brings Lenovo into 21st-century server management.

Verdict

The ThinkSystem SR650 is a versatile data centre workhorse at a great price made all the more appealing by Lenovo’s smart new remote management controller

(As reviewed)

Chassis: 2U rack

CPU: 2 x 8-core 2.1GHz Intel Xeon Silver 4110

Memory: 32GB TruDDR4 (max 3TB)

Storage: 2 x 400GB Lenovo SAS3 SFF SSD (max 24)

RAID: Lenovo ThinkSystem RAID 930-8i SAS3/2GB Flash/BBU

Array support: RAID0, 1, 10, 5, 50, 6, 60

Expansion: 7 x PCI-Express 3 (6 free)

Network: 4 x Gigabit LOM

Power: 2 x 750W hot-plug PSUs

Management: XCC Enterprise, xClarity Administrator

Warranty: 3yrs on-site NBD

Options: Dual mirrored 128GB M.2 SATA SSD kit, £328 ex VAT

HPE's ProLiant DL380 has consistently been at the top of the best seller list - and with good reason, as this is its most capable rack server. There's little it can't turn its hand to and we exclusively bring you the first hands-on review of HPE's new Gen10 flagship.

There's plenty to get excited about as along with support for Intel's Xeon Scalable CPU family, the DL380 Gen10 receives a complete refresh in the storage department with a new modular design. HPE's Smart Array RAID controllers have also been uprated to offer more storage and interface choices.

The server supports 3TB of DDR4 SmartMemory and can handle up to 192GB of HPE's Persistent Memory NVDIMMs. These target high-demand apps such as analytics or databases and provide 8GB of high-performance DRAM, backed up by 8GB of flash mounted on a standard form factor DIMM module.

And then there's HPE's new iLO 5 management controller. It offers a heap of new features as along with improved server monitoring, it has a laser-sharp focus on platform security with silicon fingerprinting, encryption and breach detection technology.

Storage

On review is the Performance model 826556-B21 which sports a pair of 12-core 2.3GHz Intel Xeon Gold 5118 CPUs teamed up with 64GB of DDR4. A quick glance at the front shows HPE has retained the 8+8+8 drive bay configuration layout, allowing up to 24 SFF drives to be presented.

You can fit HPE's universal media unit in the left bay, which provides dual USB3, a DisplayPort plus optical drive and room for two SFF drives beneath. There's more going on at the back: three dual SFF hot-swap cages can be added, taking the tally to thirty drives.

The SFF chassis also supports up to 20 NVMe drives. If capacity is a key requirement, it'll take twelve LFF drives at the front, three at the back and even four in the middle with HPE's custom mid-plane kit.

All models start with HPE's embedded Smart Array S100i chip, which supports RAID 0, 1, 5, and 10 for up to 14 SATA drives. Our server has the Smart Array P408i-a snap-in module which adds SAS3 plus 2GB of flash-protected cache to the storage equation and HPE offers other models with differing cache quantities and external expansion ports.

Management and security

The iLO 5 uses a new ASIC and its Secure Start swings into action when the server is powered on. It validates its firmware using HPE's Silicon Root of Trust fingerprinting and if it fails this, it'll stop the server booting. Next up is Secure Recovery which validates the iLO firmware and if the test fails, automatically flashes it with the resident recovery image.

Included in the iLO 5 Standard license is active system ROM validation and if this fails, it'll activate the redundant system ROM. Should both ROMs fail validation, the Advanced Premium Security Edition license can run a firmware scan and repair process and will also run daily system firmware validations.

The iLO 5 is twice as fast as iLO 4 with double the memory and we found its Intelligent Provisioning OS deployment tools and virtual media services much quicker and more responsive. The iLO 5 web interface gets a major revamp making it even more informative and HPE has switched entirely to its agentless management service (AMS), negating the need to load any OS agents.

Existing users of HPE's OneView must upgrade to v3.1 to support Gen10 hardware and patience is required, as we found this takes around two hours. With this applied, we added the server's iLO 5 credentials to the lab's OneView Hyper-V VM and kept an eye on CPU and power usage plus temperatures, ran remote control sessions and powered the server up and down.

Power and networking

There's room to grow, as our model includes both 3-slot PCI-Express riser cards with space for a 2-slot tertiary riser over the PSU bay. HPE's tool free design comes into play as the two main risers can be easily removed with just a flick of the wrist.

All models come with quad embedded Gigabit ports and you can upgrade with a choice selection of FlexibleLOM mezzanine cards. The server supports one card which fits in a dedicated slot at the rear of the motherboard and HPE offers dual and quad-port fibre or copper 10GbE plus a dual-port 10/25/40GbE card.

Our system includes a pair of 800W hot-plug PSUs and HPE offers 500W and 1600W versions, but not the Flex Slot Battery Backup Module from the Gen9 servers. For a 48 virtual core server, power consumption is commendably low with it drawing 81W in idle and peaking at 272W with both CPUs under extreme load.

Conclusion

HPE's DL380 just keeps getting better with every generation and the latest Gen10 delivers new and welcome features in abundance. This highly expandable 2U rack server offers a wealth of storage choices and the new iLO 5 controller provides top-notch remote management and platform security. Enterprises and SMBs looking for a powerful and secure 2U rack server that can grow with them will find the ProLiant DL380 Gen10 an ideal partner.

Verdict

Probably the most versatile 2U rack server on the planet, the DL380 Gen10 delivers a superb range of new features including big storage choices, slick iLO 5 management and tough security measures

Chassis: 2U rack CPU: 2 x 12-core 2.3GHz Intel Xeon Gold 5118 Memory: 64GB ECC DDR4 2,400MHz (max 3TB) Drive bays: 8 x hot-plug SFF (max 30) Storage: No HDDs included as standard RAID: HPE Smart Array S100i and P408i-a SAS3/2GB with FBWC/BBU Array support: RAID0, 1, 5, 6, 10, 50, 60, 1 ADM, 10 ADM Expansion: 6 x PCI-e Gen3 slots (max 8) Network: 4 x embedded Gigabit Other: FlexibleLOM, internal USB 2 (2), microSD card slot Power: 2 x 800W hot-plug PSUs Management: HPE iLO5 Standard with Gigabit Warranty: 3yrs on-site NBD

Intel's Purley platform promises the biggest shakeup in the server world since the 'Nehalem' launch over eight years ago and in this exclusive review, we look at the first production server to show off Intel's new Skylake-SP Xeon Processor Scalable family.

On review we have Dell EMC's PowerEdge R740xd, which heralds in its 14th Generation PowerEdge series. It showcases a completely new architecture with even more storage options, enhanced system security and a radical revamp in the server management department.

Designed to support a wide range of business applications including VDI and software defined storage, the PowerEdge R740xd allows you to mix and match any type of drive in its chassis. With support for HDD, SSD, boot-optimized M.2 SSDs and up to 24 NVMe SSDs, it allows businesses to tailor storage precisely to their capacity and performance requirements.

Specs and hardware

The Xeon Processor Scalable family is enormous, with Bronze, Silver, Gold and Platinum products targeting the performance, energy-optimized, 10-year extended life and Omni-Path Fabric sectors. You have two Bronze 31xx models, eight 41xx Silvers, six 51xx Golds, twenty-three 61xx Golds and twelve 81xx Platinums.

All use the same LGA-3647 Socket P - so you can forget all about Xeon DP, MP, E5, E7 and so on, as these have now been merged into one product line. Bronze, Silver, Gold and Platinum CPUs all support dual socket deployments, Gold and Platinum support 4P configurations while Platinum supports 8P.

Our review system came with dual 2.1GHz Intel Xeon Gold 6130 CPUs delivering an impressive 32 physical and 64 logical cores. A big bonus of the Socket P implementation is that the R740xd supports all models, right up to the massive 28-core 2.5GHz 8180 Platinum Xeon.

All have six DDR4 memory channels, but only the Gold 5122, 61xx and Platinum CPUs support the faster 2,667MHz variety. Our server came supplied with 192GB of 2,667MHz memory, and its 24 DIMMs slots allow this to be expanded to 3TB.

Cooling technology

Cooling for this 2U chassis has been completely redesigned, with the more open honeycombed front bezel improving air flow. The server is available with 12 LFF or 24 SFF hot-swap front bays and the carriers have also been redesigned for better air flow.

It's no wonder Dell EMC had to redesign cooling, as this server is jam-packed with internal features. Six hot-plug fans link up with a multi-vector cooling feature, which senses when certified PCI-Express cards are installed and automatically adjusts the air flow to them.

A removable full-width tray takes four NVMe SSDs. Underneath this are two CPU sockets, fitted with big passive heatsinks and flanked by 24 DIMM slots. If you plan of expanding to 3TB of memory you must specify M-designated Gold or Platinum CPUs, as only these support 1.5TB of memory per socket.

Storage

RAID options haven't changed much from the 13th Gen PowerEdges, where base systems start with an embedded PERC S140 chip supporting SATA drives plus stripes, mirrors and RAID5 arrays. Dell EMC offers a wide range of PERC 10 RAID controllers, and our system sported a PERC H730P Mini card with 2GB of NVRAM cache and integral battery backup.

There's more on the storage front, as the R740xd has another drive tray at the back which supports a further 2 LFF or 4 SFF hot-swap drives. The cage is cabled to the front backplane and integrates directly with the resident RAID controller.

The NVMe cage requires a PCI-Express extender card which slips into a slot beneath the rear drive tray. The R740xd has even more levels, as under this slot is a connector for its network daughtercards.

We tested the dual Gigabit/10GBase-T model, and even this had dedicated copper heat-pipes and radiators for cooling. Other network options are plentiful and include quad Gigabit, quad 10GbE or dual 25GbE.

Management features

The iDRAC9 controller uses a completely new chip, which provides a slick HTML5-based web interface. This is vastly superior to the iDRAC8 and, dare we say it, even better that HPE's iLO4.

The dashboard provides a complete status overview, while the System tab presents a wealth of operational data about areas such as batteries, cooling and voltages. It offers a full system inventory and performance graphs, plus direct access to the PERC for RAID configuration.

Data centres will love the new Quick Sync 2 feature. Pressing the button under the system ID LED on the front panel activates Bluetooth, and allowed us to connect our iPad to the server via the OpenManage Mobile (OMM) iOS app. Using OMM to scan the QR code on the system label, we had the R740xd loaded in seconds where we could view all server information, alerts and health status, plus launch a virtual console for remote control.

Security is extensive, as the iDRAC9 System Lockdown mode stops users changing any configuration settings. On top of that, along with full supply chain security, firmware packages are cryptographically signed and storage can be safely wiped using the iDRAC9 System Erase feature.

We could also monitor the R740xd via the lab's OpenManage Essentials host, which provided hardware and software inventories along with advanced alerting features. OpenManage Enterprise will be the next step but Dell EMC was unable to advise us when this would be released.

Verdict

As one of the first Skylake-SP servers to market, the PowerEdge R740xd sets a very high standard for the rest of the blue chips to follow. The superb internal design allows its 2U chassis to offer an unbeatable range of storage options and the new iDRAC9 delivers excellent remote management tools. It's the perfect choice for data centres that want maximum processing power and storage performance in their rack cabinets.

Verdict

Along with early support for Skylake-SP Xeons, the PowerEdge R740xd packs a superb range of storage features into its 2U chassis and teams them up with a class remote management act

Fujitsu is justifiably proud of its new 2U rack server. Not only has the Primergy RX2540 M2 claimed four performance world records for SAP, VMmark and TPC-E benchmarks, but its versatility makes it ideally suited to a wide range of business applications.

SMBs can start small and upgrade as demand increases. Supporting E5-2600 v4 Xeons, Fujitsu offers 21 options, allowing businesses to pick and choose the CPUs for their budget and workloads.

It supports up to 24 SFF drives, which can be increased to 28 with an optional rear cage. Flexibility is the watchword too, as Fujitsu offers a range of expansion kits allowing you to upgrade capacity as required.

The RX2540 M2 has a very clean interior, making upgrades a swift process

Storage spaces

Base systems start with the 4-port embedded SATA controller plus support for stripes and mirrors, and the server can be expanded massively beyond this. The choice is yours as the front panel supports 4, 8 and 12 LFF drives or 8, 16 and 24 SFF drives.

Our system came with 16 SFF bays with the first eight cabled directly to a PRAID EP420i PCI-e SAS3 controller card with 2GB of cache. To use the second group of eight bays, we needed to order an upgrade kit comprising backplane plus SAS expander and re-cable our RAID card.

The same applies to the 24-drive chassis, although you lose the operator panel and 5.25in bay underneath. Either way, systems can be field upgraded as Fujitsu offers LFF and SFF upgrade kits that cover every eventuality.

The extra four-bay SFF cage is fitted at the rear above the PSU bays. This arrangement stops the leftmost PCI-e slot being used, though, and the bay must be connected to a dedicated RAID card.

Add the optional 4-bay SFF cage at the back and you can push the drive count to 28

Cooling and ATD

Cracking the lid reveals a well-designed interior with all components easily accessible for routine maintenance. Five hot-plug fans handle cooling, with the CPU and memory sockets covered by a solid transparent air shroud.

During testing we found the fans wouldn't settle at less than 8,000rpm, which resulted in slightly high noise levels of 67dB at one metre. The server doesn't support Fujitsu's LNM (low noise mode) feature, which is only available on its older rack systems.

Even so, ATD (advanced thermal design) enabled servers are certified to run in ambient temperatures up to 45 degrees C, allowing data centres to reduce air-con running costs. ATD is a factory option but what does it involve?

BIOS mods are made to adjust fan sensor thresholds so they kick in earlier. Air flow through the chassis must be unimpeded so ATD is only available on systems with 4 LFF or 8 SFF drives, while other restrictions include high TDP processors, GPU cards, internal tape drives and RAID flash cache modules.

We used the iRMC web interface to monitor the server's power consumption

Management news

Remote management comes up short when compared with HPE and Dell. The embedded iRMC S4 chip offers a dedicated Gigabit port and web interface that provides plenty of data on critical components.

It isn't as sophisticated as HPE's slick iLO4 interface but help is on the way. Fujitsu is planning a major iRMC upgrade to coincide with Intel's launch of its next-generation Purley' CPUs.

Its ServerView Suite software also needs a spring clean, as although it provides plenty of utilities for server monitoring, it is dated. This is also expected to see a major refresh this year in-line with the iRMC updates.

The elderly ServerView software uses an agent to remotely monitor the server's vital signs

Processing and power

Our review system included dual 2.2GHz E5-2650 v4 Xeons teamed up with 64GB of DDR4 memory. The motherboard has 24 DIMM slots with maximum memory currently at 1.5TB, although those with deep pockets can push this to 3TB using 128GB LRDIMMs.

The system on review includes a quartet of 300GB 15K SAS3 SFF drives and the PRAID EP420i controller. You can reduce costs by opting for the CP400i or EP400i SAS3 cards. These support RAID5 and 50 arrays with the latter including 1GB of cache memory.

Power choices are plentiful as our system has dual 450W hot-plug PSUs, while more powerful systems can avail themselves of 800W Platinum or Titanium PSUs or 1200W Platinum versions. Even with the fans running faster than we'd like, the server returned good readings in our power tests, consuming 119W in idle and peaking at 282W with the CPUs under a 100% load.

The server supports up to eight PCI-e cards using risers and there are also Fujitsu's DynamicLOM adapters. Snapping into a proprietary slot on the motherboard, versions are available with dual or quad Gigabit and dual fibre or copper 10-Gigabit.

Boot the system with the ServerView disk and it'll help you load an OS

Conclusion

With a list price of 5,000 for the review system, the Primergy RX2540 M2 looks very good value (street prices will almost certainly be lower). This includes a fine hardware package, making our review system ready for some serious app bashing straight out of the box.

It's a solid investment as the server can be specified to suit your budget and easily upgraded as required. Storage won't be as problem either, as the RX2540 M2 offers upgrade kits that cover most future business requirements.

Chassis: 2U rack CPU: 2 x 2.2GHz Xeon E5-2650 v4 Memory: 64GB DDR4 (max 1.5TB) Storage: 4 x 300GB HGST 15K SAS SFF (max 28) RAID: Fujitsu PRAID CP420i 12Gbps SAS/2GB cache Array support: RAID0, 1, 1E, 10, 5, 50, 6, 60 Network: 4 x Gigabit DynamicLOM card Expansion: 6 x PCI-e 3.0 (max 8 with risers) Power: 2 x 450W hot-plug PSUs Management: Fujitsu iRMC S4 Advanced with Gigabit Warranty: 3 years on-site NBD

The 2U rack server is the undisputed workhorse for most businesses. Lenovo claims its System x3650 M5 is its most flexible yet with a wide range of new features including improved performance, greater energy efficiency and even more storage options.

SMBs who want a server that can grow with their needs will like the x3650 M5 as it provides a highly flexible upgrade path. It supports the latest E5-2600 v4 Xeon processors plus 1.5TB of TruDDR4 memory while storage can be increased easily.

Lenovo has also improved its infrastructure management tools in the form of the new xClarity software portfolio. In this exclusive review, we put the xClarity Administrator through its paces and see how it measures up.

Storage decisions

Lenovo offers an intense amount of storage options. Base systems start with eight SFF or eight LFF drive bays. The SFF models use a front panel with an 8+8+8 configuration so you can add blocks of eight drives as required.

RAID choices are equally flexible. All three SFF backplanes can be linked together and managed by the ServeRAID M5210 controller or split up and handled separately by one or two additional cards.

The eight-drive LFF models still have room for Lenovo's operator panel above them. Alternatively, you can dispense with this and cram twelve LFF bays into the front.

Along with support for eight NVMe PCIe SSDs and four GPUs, the server has space at the back for even more drives. Depending on your main configuration, you can add two dual-drive SFF cages, one dual-drive LFF cage or a mixture of both.

Tidy design

The server's internal design makes light work of upgrades and routine maintenance. There's tons of room to expand as the motherboard has two embedded PCI-Express slots and also accepts a pair of three-slot riser cards.

The server has plenty of room to expand, plus the ServeRAID card has its own dedicated slot

The ServeRAID M5210 card doesn't waste space either as it's fitted in a ninth horizontal PCI-Express slot behind the CPU and memory sockets. It supports 12Gbps SAS 3 drives and offers upgrade paths including RAID5/50, RAID6/60 and a choice of 1GB, 2GB or 4GB caches with flash backup.

Cooling is handled efficiently by a bank of six hot-swap fans and we found the server to be quiet with the SPLnFFT iOS app measuring only 49dB from one metre away. Lenovo's calibrated vectored cooling splits them into two zones allowing the server to operate in ambient temperatures up to 40 degrees Celsius.

The CPU and memory sockets are covered by a sturdy yet easily removable air shroud. The TruDDR4 memory has built-in signatures that the server will authenticate. Once this is done, Lenovo claims they'll run up to 12 percent faster than Intel's specifications when placed in two RDIMM per channel configurations.

xClarity brings network clarity

A legacy from this server's IBM days is its embedded IMM 2 remote management controller. While its web console does provide plenty of information about critical components, it can't match Dell's iDRAC8 or HPE's iLO4 for features.

The embedded IMM2 controller only has a basic web interface

The new xClarity Administrator remedies this as it can manage your entire Lenovo infrastructure including System x and ThinkServer systems, switches, storage arrays and rack cabinets. Even better, it doesn't require agents and we had the Hyper-V version loaded and running inside 30 minutes.

A quick network discovery found our server, inventoried it and added it to the server pane in the dashboard. This provides an overview of managed systems neatly grouped into types and accompanied by coloured status icons.

Lenovo's classy xClarity Administrator shows how serious it's taking enterprise server management

We monitored critical hardware components, viewed power consumption for everything including the memory and processors and ran remote control sessions. xClarity also teams up an OS image library and firmware repository with automated provisioning tools.

The image library does away with the need to use Lenovo's ServerGuide DVD to deploy an OS. Enterprises will also approve of the xClarity server patterns as these define complete server configurations and are used to set up or repurpose multiple servers in one go.

From the xClarity console we viewed the server's vitals and fired up remote control sessions

Power and price

Our review server came with a 750W Platinum PSU and there's room for a second. Lenovo offers a wide choice with 550W, 900W, 1300W and even 1500W versions available. Power efficiency is impressive as we measured the system drawing 80W in idle and peaking at only 167W under extreme load.

The system on review here was generously equipped with a pair of eight-core 2.1GHz E5-2620 v4 Xeons, a plentiful 128GB of TruDDR4 memory, the ServeRAID M5210 controller and two 120GB SATA SSDs. SMBs that want to start smaller will be tempted by the entry-level 8871KAU model, available through Lenovo's TopSeller program, which comes equipped with a 1.7GHz six-core E5-2603 v4 Xeon plus 16GB of TruDDR4 and costs only 1,400.

Server power consumption and thermal values can be viewed remotely from the xClarity console

Conclusions

The System x3650 M5 shows Lenovo is getting very serious about its servers as it has an impressive specification for the price. It's also one of the most flexible servers we've yet seen and its multitude of storage upgrade paths make it a worthy long-term investment.

Power efficiency is very high while build quality and design is near faultless. The icing on the cake is Lenovo's slick xClarity Administrator which provides an easy way of managing your entire network.

Verdict

A powerful, highly expandable 2U rack server made even better with Lenovo’s new server management software

Chassis: 2U rack

CPU: 2 x 2.1GHz Intel Xeon E5-2620 v4

Memory: 128GB 2,400MHz TruDDR4 (max 1.5TB with 64GB LRDIMMs)

Drive bays: 8 x SAS/SATA SFF (max 28)

Storage: 2 x 120GB SFF SSDs

RAID: Lenovo ServeRAID M5210 12Gbps SAS3

Array support: RAID0, 1, 10 (5, 6 optional)

Expansion: 9 x PCI-E Gen3 (with two risers)

Network: 4 x embedded Gigabit Ethernet

Power: 750W Platinum hot-plug PSU (max 2)

Management: IMM2 Advanced

Warranty: 3 years on-site NBD

Options: Lenovo xClarity Administrator - £141 ex VAT per managed server/year

Boston's Fenway 2224-0T is a server aimed at businesses that want something of everything. This 2U rack system combines the latest E5-2600 v4 Xeons with fast 2,400MHz DDR4 memory, soups it up with a healthy dose of NVMe SSD storage and delivers it all at a sensible price.

It's an all-Supermicro system comprising its SuperServer 6028U-TNRT+ rack platform. You don't get to choose your motherboard as the chassis is specifically designed for the X10DRU-i+ model with Supermicro supplying the system fully assembled.

This is because the four front drive bays to the right are dedicated to NVMe SSD storage with orange markers on the carriers to denote this. The drive backplane has four dedicated connectors cabled to a riser card at the rear. The riser card is plugged into a row of proprietary PCI-Express slots on the motherboard.

The system is also only available with two processors installed. The reason is quite simple as it uses CPU1 to activate the NVMe riser card and CPU2 to enable the standard PCI-Express riser cards across the way.

The system requires both CPUs installed to enable the NVMe ports and all PCI-Express slots

Storage connections

The Fenway has plenty of storage permutations as, along with the four NVMe SSD bays, its SAS3-ready backplane services all eight standard hot-swap drive bays. The motherboard has ten embedded 6Gbps SATA3 ports and its chipset supports mirrors, stripes and RAID5 arrays.

Boston improves your storage outlook further by including a Supermicro AOC-S3108L-H8IR RAID card. This delivers 12Gbps SAS3 support plus RAID6 to the first eight drive bays and has 2GB of onboard DDR3 cache.

To keep costs down, Boston chose four 6TB WD Enterprise SATA drives to provide a high-capacity data storage repository. The price includes a 400GB Intel DC P3500 MLC NVMe SSD for the OS and our tests confirmed that it is exceedingly fast with Iometer reporting 4KB random read and write rates of 282,000 and 50,000 IOPS.

Under the lid

We aren't fans of Supermicro's button release latches on its rack server lids. These can be very stiff making it difficult to remove the lid this was doubly annoying when the server was dangling out the front of our rack cabinet on fully extended rails.

However, once inside we found the well-designed interior afforded easy access to all components. Our review system included a pair of 2.2GHz E5-2650 v4 Xeons with twelve physical cores apiece, a fat 30MB L3 cache, a 9.6GT/sec QPI speed and the ability to turbo boost core speeds to 2.9GHz in times of great need.

The price includes 64GB of 2,400MHz DDR4 memory and with 24 DIMM slots up for grabs, there's plenty of room for more. The motherboard supports up to 768GB of RDIMM memory and a whopping 1.5TB using more expensive LR-DIMMs.

Expansion and power

With the two processors in residence, all seven PCI-Express slots are activated. These comprise five full-height plus two low-profile versions - the three cages are easily removed simply by pulling out a small locking tab behind each one.

The embedded RMM provides a simple web interface for remotely managing the server

Cooling is handled efficiently by four hot-swap fan modules, but we were disappointed by the flimsy plastic air shrouds covering the CPUs. Even so, they work well enough as air directors. The fans were also quiet with the server emitting 56dB on their optimal setting and only 51dB with the PUE2 (power usage effectiveness) mode set.

Power won't be a problem either as the Fenway is endowed with two 1kW Titanium Plus hot-plug PSUs. Even though the Xeons have a 105W TDP, we clocked the system pulling 186W in idle and peaking at a very acceptable 285W under the load of our benchmarks. For comparison, HP's ProLiant DL380 Gen9 and its dual 14-core 2.3GHz E5-2695 v3 Xeons hit 352W under load.

Server management

Along with dual embedded 10GBase-T data ports, the server has a dedicated Fast Ethernet port for remote server management. The web interface isn't a patch on Dell's iDRAC8 or HP's iLO4, but it does provide valuable sensor information on key components and the facilities to link thresholds to SNMP traps and email alerts.

The RMM provides information on all key sensors which can be linked to alerts

It provides historical graphs of power consumption, a separate screen for viewing details of installed NVMe SSDs and remote power controls are provided too. Supermicro scores extra points for value as the controller includes full OS remote control and virtual media services as standard and not, as with HP and Dell, as expensive upgrades.

Surprisingly, Supermicro includes full OS remote control at no extra cost

Supermicro also includes its free SuperDoctor 5 SNMP Windows management utility. Using a web browser, we could access the server locally or remotely and view a gaudy graphical display of fan speeds, voltages, temperatures, disks and memory.

The SuperDoctor SNMP app provides a cheery dashboard full of coloured graphs and speedo dials

Conclusions

Boston's Fenway 2224-0T is great value as equivalent systems from some of the blue chips will set you back at least 1,000 more. It combines the latest Xeon E5-2600 v4 processing power with SAS3 support plus fast NVMe SSD storage and, apart from a couple of minor build quality issues, is well designed with plenty of headroom to handle demanding business apps.

Verdict

Boston delivers a powerful and affordable 2U rack server package with Xeon E5 v4 CPUs, fast NVMe SSD storage and plenty of opportunities to expand

Chassis: 2U Supermicro SuperServer 6028U-TNRT+

Motherboard: Supermicro X10DRU-i+

CPU: 2 x 2.2GHz E5-2650 v4 Xeon

Memory: 64GB 2,400MHz 1.2V DDR4 RDIMM (max 1.5TB with LR-DIMMs)

Storage: 400GB Intel DC P3500 NVMe (max 4), 4 x 6TB WD 6Gbps SATA HDDs (max 8)

RAID: Supermicro AOC-S3108L-H8IR SAS3 PCI-Express card/2GB DDR3 cache

Array support: RAID0, 1, 10, 5, 50, 6, 60

Expansion: 7 x PCI-Express Gen3 slots

Network: 2 x embedded 10GBase-T

Power: 2 x 1000W Titanium Plus hot-plug PSUs

Management: Supermicro RMM with 10/100 port

Warranty: Three years on site NBD

Gigabyte may not have the same kudos as Dell or HP in the server market, but it does offer a good range of affordable if more basic servers. Broadberry's latest Gigabyte-designed CyberServe XE5-R280 2U rack server is nonetheless one of the first to be based around Intel's new Xeon E5-2600 v4 Broadwell-EP' processors and we've got an exclusive look at it.

Designed for two-socket systems, the new Xeons are a tick' in Intel's server CPU development cycle and have more cores than the previous generation. The top-end 2.2GHz E5-2699 v4 Xeon, for example, has 22 cores and a stonking great 55MB L3 cache.

The family is suited to a wide range of business applications and, with 21 server models to choose, you can pick the right ones for your workload. Along with a fab drop to 14nm, Intel has added a heap of performance enhancements including 2,400MHz DDR4 memory support, advanced virtualisation capabilities, memory bandwidth monitoring, accelerated AES encryption and much more.

Build quality

To showcase the new Xeons, Broadberry chose Gigabyte's R280-A3C 2U rack chassis and MD90-FS0 motherboard. It's a good combination as this partnership offers high memory and storage capacity plus big expansion potential and it's certified to support the high-performance 160W E5-2687W v4 Xeon.

The chassis is well-made with the twelve hot-swap LFF drive bays linked up to a SAS3 backplane with integral SAS expander. Behind this is a bank of four cold-swap cooling fans underneath a separate removable panel.

The motherboard is easily reached by removing the rear panel and we found access to all key components was unhindered. Although the MD90-FS0 board was originally launched for the E5-2600 v3 Xeons, it just needed a BIOS upgrade to support the v4 E5 Xeons as the new processors are simple drop-in upgrades.

Storage galore

The CyberServe can't be faulted for storage features our review specification includes a great package for the price. We asked for it to be supplied with Windows Server 2012 R2 and Broadberry loaded it on a quartet of 800GB Toshiba SAS3 eMLC SSDs configured in a RAID-10 array.

OS performance won't be a concern as these SSDs are fast. Running Iometer with 4KB transfer requests, we recorded high random read and write throughputs of 140,000 and 38,000 IOPS respectively.

The price also includes a big data repository comprising four 6TB Toshiba SAS3 LFF drives configured in a RAID5 array. This is very generous as street prices are around 4,400 for all eight drives whereas Broadberry is charging nearly 500 less for this configuration.

The drives are managed by an Intel RS3DC080 SAS3 PCI-Express card cabled directly to the drive backplane. Along with support for all the usual RAID suspects including RAID6, it comes as standard with 1GB of 1,866MHz ECC DDR3 cache and the price includes the optional battery backup unit.

The server chassis is well designed and provides an impressive expansion potential

Design and power

Broadberry supplied the CyberServe with two 2.4GHz E5-2640 v4 Xeons for a total of 20 physical cores. Although the motherboard supports 2,400MHz DDR4 memory, these CPUs max out at 2,133MHz and have a top QPI speed of 8.0GT/sec.

Broadberry has kept a keen eye on costs but still managed to include 128GB of 2,133MHz 1.2V DDR4 RAM. With 24 DIMM slots available, memory can be expanded to 768GB, and if you can afford 64GB LR-DIMMs, this can be pushed to an app busting 1.5TB.

The server includes dual 800W Platinum rated hot-plug PSUs and we barely stressed them. We measured the server drawing 167W in idle and peaking at only 310W when handling the load of our benchmarks.

The server's BMC includes OS remote control as a standard feature

Expansion and management

Expansion potential is impressive as the server has three riser cards giving you a total of eight PCI-Express slots with six supporting full-height, full-length cards. There's more - below the central riser is a mezzanine slot on the motherboard for which Gigabyte offers dual-port 10GbE and Infiniband cards.

You'll need a screwdriver to get at them all though, as Gigabyte doesn't yet appreciate the benefits of HP's and Dell's simple tool-less server designs. We also found the mezzanine slot difficult to access as the riser card above it must be removed at the same time.

For remote server management, the motherboard's Aspeed AST2400 BMC chip provides plenty of information about critical components along with full OS remote control tools and direct access to BIOS upgrades. The server's cooling fans were emitting a noisy 72dB on the Energy Saving' mode but we used the BMC console to manually slow them down to 57dB and keep a close eye on CPU temperatures at the same time.

The server's embedded Aspeed BMC provides plenty of details on critical components

During testing, we had to tweak the cooling fan speeds to reduce noise levels

Gigabyte's monitoring tool amalgamates information from the BMCs on multiple servers into a single console. This provides plenty of operational data and access to power controls, but it looks very rough and could be much better designed.

Gigabyte's multi-server management console is informative but isn't pretty to look at

Conclusions

This all-Gigabyte rack server has a couple of rough edges, but nonetheless scores highly for its blistering performance, fast SAS3 storage and plentiful expansion options. It's a great value server.

Verdict

The CyberServe XE5-R280 teams up the latest E5-2600 v4 Xeons with a mighty storage package and sets a high standard for value

Chassis: Gigabyte R280-A3C 2U rack

Motherboard: Gigabyte MD90-FS0

CPU: 2 x 2.4GHz E5-2640 v4 Xeon

Memory: 128GB 2,133MHz 1.2V DDR4 RDIMM (max 1.5TB with LR-DIMMs)

Storage: 4 x Toshiba 800GB SAS3 eMLC SSD, 4 x Toshiba 6TB SAS3 HDD (max 12)

RAID: Intel RS3DC080 SAS3 PCI-Express card

Array support: RAID0, 1, 10, 5, 50, 6, 60

Expansion: 8 x PCI-Express Gen3 slots, mezzanine slot

Network: 2 x embedded Gigabit Ethernet

Power: 2 x 800W Platinum hot-plug PSUs

Management: Aspeed AST2400 with Gigabit Ethernet

Warranty: 3 years on site NBD

Options: Dual-port 10GbE SFP+ mezzanine card, £220 ex VAT

Samsung has introduced a 15TB solid-state drive (SSD), which it claims is the largest drive of its kind in the world today.

The PM1633a drive is intended for enterprise storage systems, and utilises a 12GB/s serial attached SCSI (SAS) interface.

Its 15.36TB of space is equivalent to approximately 300 Blu-ray discs or 4,000 standard layer DVDs, while Samsung said its 2.5-inch form factor means enterprise storage managers can fit twice as many of the drives in a standard 19-inch, 2U rack, compared to an equivalent 3.5-inch storage drive.

"To satisfy an increasing market need for ultra-high-capacity SAS SSDs from leading enterprise storage system manufacturers, we are directing our best efforts toward meeting our customers' SSD requests," said Jung-Bae Lee, senior VP of the memory product planning and application engineering team at Samsung Electronics.

Samsung also claimed the drive can perform random read and write speeds of up to 200,000 and 32,000 IOPS respectively. On sequential read and write speeds it can apparently reach up to 1,200MB/s.

The South Korean manufacturer had first unveiled the PM1633a at the Flash Memory Summit in August 2015, saying that the drive is a product of Samsung's 3D V-NAND chip technology, and therefore should be more reliable than its predecessors.

Samsung has not yet unveiled the price or release date of the new drive, which will go on sale alongside smaller storage drives later this year, including 7.68TB, 3.84TB, 1.92TB, 960GB and a 480GB drive.

Small businesses that thought rack servers sporting the latest core-heavy E5-2600 v3 Xeons were beyond their budget should think again. HP's ProLiant DL80 Gen9 delivers these and more with prices for the entry-level models sneaking in on the right side of 1,000.

As with all its Gen9 servers, HP's has taken a different approach to system design. Although not as versatile as the flagship DL380 Gen9, the DL80 Gen9 is still a highly scalable system that allows businesses to pay only for what they need now but with a view to upgrading on-demand in the future.

The system on review is the entry model of this new family and has a tempting price tag of only 989. This gets you a single 1.6GHz E5-2603 v3 CPU with 6 cores but a second socket is ready and waiting and HP offers 9 other models if you fancy more speed and cores.

Naturally, high-speed DDR4 memory is supported and HP has cut costs by reducing the number of DIMM slots to 8. The review system came with a single 4GB stick but with dual CPUs installed, it supports up to 256GB easily enough for most small businesses.

The DL80 Gen9 supports two Xeon CPUs but DIMM slots have been reduced to eight to cut costs

Storage options

Storage choices are equally enticing as this 2U chassis can handle up to 12 hard disks. Our entry-level system has 4 cold-swap drive bays which can be upgraded to a maximum of 8 with HP's enablement kit.

RAID starts with the embedded Dynamic Smart Array B140i controller which provides dual 4-port SATA connectors on the motherboard. We could create stripes, mirrors plus RAID5 arrays and manage them easily using HP's Smart Storage Administrator utility.

HP offers configure-to-order (CTO) models if you want 8 or 12 hot-swap drive bays and its SmartDrive carriers where each provides an integral array of hard disk status LEDs. There's also a big choice of SATA and 12Gbps SAS Smart Array controllers but bear in mind you can't upgrade from 8 bays to 12 bays and must order the higher capacity model at the outset.

HP offers a good range of RAID controller options and all can be managed from the Smart Storage Administrator

Network and power

Internal design is excellent with the base system including two cold-swap fans positioned in front of the occupied CPU socket. There's room for two more on each side of these which you'll need when you upgrade to dual CPUs and use the expansion slots over the PSU bay.

All models in the family have two Gigabit ports and HP offers plenty of network upgrades including 10GbE, Fibre Channel and more. Even with only a single CPU in residence, we had five PCI-Express slots available so upgrade potential is excellent.

With two CPUs installed, we could add additional riser cards to increase the slot count to 7 and there's room over the PSU bay for a GPU card. The standard 500W cold-swap PSU won't be enough for a GPU but you can upgrade to 800W redundant PSUs.

The E5-2603 v3 and its low 85W TDP makes the DL80 Gen9 a real power miser. In idle, we clocked the review system drawing a paltry 45W and peaking at only 71W when under maximum load from the SiSoft Sandra benchmarking app.

HP's iLO4 controller delivers unbeatable remote server management and monitoring tools

iLO4 included

Server management doesn't get any better as the server is blessed with HP's iLO4 controller which provides a wealth of information about server operations plus access to remote support. HP's larger servers provide a dedicated port for network access but on the DL80 Gen9, the iLO4 controller shares access with the first Gigabit port.

From the iLO4 web interface, we could monitor all critical components and remotely control the server's power. However, you'll need an Advanced license to get features such as OS remote control, virtual media, remote console record and playback plus advanced power monitoring and dynamic power capping.

OS installation took us less than an hour using the embedded Active Intelligent Provisioning tool which does away with the need to boot the server with a deployment disk. We also loaded the free System Management HomePage utility which provides a browser interface offering basic local monitoring of critical components.

HP's Insight Control handles general network systems management. We also used HP's iOS Insight Online app with our Passport account to view registered systems, monitor support cases and check support contracts.

HP has tightened up its server support as you only get access to regular ROM BIOS firmware updates with a valid warranty or support contract. However, it will provide open access to critical firmware updates regardless of support agreements.

HP's iOS Insight Online app allowed us to access our Passport account, view support contracts and download documentation

Conclusion

HP's ProLiant DL80 Gen9 is a fine choice for storage-focused SMBs that want a rack server that can grow with them at a price they can afford. There are a lot of storage options so choose carefully when ordering but its scalable design makes it a good long-term investment and its E5-2600 v3 Xeon power and low price earns it a well-deserved Recommended award.

Verdict

SMBs that want the latest E5-2600 v3 Xeons in their server will like the budget-friendly ProLiant DL80 Gen9. It may have a low starting price but features aren’t compromised and its high capacity makes it a fine choice for storage-centric apps

Chassis: 2U rack CPU: 1.6GHz E5-2603 v3 Xeon (max 2) Memory: 4GB 1,600MHz DDR4 (max 256GB) Storage: No HDDs included (max 12 LFF with CTO) Drive backplane: Cold-swap 4-bay backplane RAID: HP Dynamic Smart Array B140i Array support: RAID0, 1, 10, 5 Expansion: 7 x PCI-e 3 (with 2 CPUs) Network: 2 x Gigabit Other: Embedded USB 3 port and SD Card slot Cooling: 2 x cold-swap fans (max 6) Power: 500W cold-swap PSU (max 2) Management: HP iLO4 Standard Warranty: 1yr on site NBD

Dot Hill's innovative design allows the Ultra48 to pack more storage into less space than any competing devices. Instead of spreading the drives across the front panel in the traditional manner, the Ultra48 employs three extendable drawers each with 16 SAS 2 hot-swap SFF bays.

The drawers can be pulled out without disturbing existing RAID arrays allowing more drives to be added on the fly. All the bays face inwards allowing full access even if the system is in a rack at the end of a row next to a wall.

Clever design doesn't stop there as the Ultra48 controllers support both 16Gbps FC and 10GbE operations (there's also a version with 12Gbps SAS host ports). The four ports on each controller can be configured to accept 16Gbps FC or 10GbE SFP+ transceivers or a mix of both.

Redundancy is total as the chassis has two hot-plug power supplies with integral fans and the dual controllers run in active/active mode with cache mirroring. Each controller has 6GB cache memory and uses Dot Hill's EcoStor combo of super-capacitor and CompactFlash card to protect the cache contents.

Storage is spread across three extendable drawers with each offering 16 hot-swap SFF bays

Deployment and vdisks

The RAIDar web interface is well designed and provides easy access to all storage functions. It offers a heap of graphical views of the front and rear of the system plus a map showing which bays are occupied.

Vdisks (virtual disks) are used to provision storage where you select the member drives and decide on a RAID array type. During creation you can step through choosing member drives, the number of volumes it should contain, their size and which data ports they should be mapped to for host access.

You can add hot-spares to selected arrays and if vdisk space gets tight you can expand them into available drives. However, Dot Hill still doesn't support migration from one RAID array type to another.

The RAIDar web interface provides a handy graphical map to keeps tabs on all those drives

AssuredSnap and AssuredCopy

DotHill's AssuredSnap feature creates point-in-time snapshots of selected volumes which are stored in separate pools on the array where they can be used for fast restores. The price we've shown doesn't include snapshots but you can apply a 60-day trial license from the web interface and have a play with them. The Ultra48 supports up to 1,024 snapshots and enabling these costs a further 1,516.

You can create snapshots during vdisk configuration, run on-demand snapshots manually or schedule them to run at regular intervals. Snapshots of multiple volumes can be created quickly simply by selecting the vdisk they're contained in.

Dot Hill's AssuredCopy creates identical copies of volumes which can then be brought online for use in testing or data archiving. Volume copies can be run immediately, scheduled for later on or run at regular intervals. AssuredCopy is optional feature but enabling it to support 1,024 copies only costs an extra 640.

Each controller has four SFP+ host ports and here we've configured them for dual 16Gbps FC and 10GbE IP SAN operations

Expansion and host port options

The controllers each have a 6Gbps SAS expansion port for daisy-chaining disk shelves and with dual controllers, you can use multiple paths for fault tolerant links. The Ultra48 supports three 48-drive expansion shelves for a total of 192 drives.

Configuring the SFP+ host ports is a simple task as one CLI command determines their mode. You can have all eight ports on both controllers operating as 16Gbps FC or as 10GbE IP SAN.

Alternatively, you can choose a mixed mode where the first two ports on each controller are 16Gbps FC and the second pair are 10GbE. We remotely accessed the array over SSH using PuTTY and changed the mode in seconds.

The Ultra48 supports up to 1,024 snapshots for fast point-in-time copies of volumes

Performance

For testing we used a Broadberry Data Systems rack server with dual 2.6GHz E5-2670 Xeons, 48GB of DDR3 and running Windows Server 2012 R2. We added a QLogic QLE2562 dual-port 8Gbps FC HBA and used a direct connection to the Ultra48 array.

Raw sequential read performance over 8Gbps FC was good. We used Iometer with a 50GB test file and watched it record a steady speed of 785MB/sec or 6.1Gbits/sec. Raw write performance wasn't so hot with our test server returning 440MB/sec. To test maximum IOPS we used a 4KB Iometer transfer request and saw around 34,000 IOPS.

We then added a second Broadberry dual Xeon E5-2400 server also equipped with a QLogic 8Gbps FC card. With both servers running Iometer on dedicated volumes in the same RAID-5 array, we recorded a cumulative raw sequential read speed of 1490MB/sec a tidy 11.6Gbits/sec.

We tested over 10GbE and reconfigured the host ports accordingly. With our dual 2.6GHz E5-2670 Xeon system connected via its Intel 10GbE port, Iometer returned an impressive sequential read rate for a 1.5TB target of 963MB/sec. With an Iometer 4K read request set we also recorded a high 56,000 IOPS for the target.

Conclusion

The AssuredSAN Ultra48 and its innovative design sets new standards for rack storage density. It offers high-performance and for a dual 16Gbps FC, 10GbE IP SAN appliance it's good value too.

Verdict

No other vendor can pack this many drives into a 2U chassis. Dual support for 16Gbps FC and 10GbE IP SANs makes the Ultra48 versatile and easy to deploy.

Chassis: 2U rack

Storage: 48 x 6Gbps SAS SFF hot-swap drive bays

Power: 2 x 928W hot-plug PSUs

Two hot-swap controllers each with the following:

Memory: 6GB cache with super-capacitor and CompactFlash card

RAID support: RAID0, 1, 3, 5, 6 10, 50

Data ports: 4 x 16Gbps FC/10GbE SFP+

Management: 10/100 Ethernet

Expansion: 6Gbps SAS port

Other ports: Mini-USB CLI port

Management: Web browser

Options: AssuredSnap (1,024 snapshots), £1,156; AssuredCopy (1,024 copies) £640 (all ex VAT)

As a general purpose rack server, Dell's PowerEdge R720 has always been a top choice and this latest version offers an even higher processing density and increased performance. It supports Intel's E5-2600 v2 CPUs which deliver more cores and better performance without a significant increase in power demands.

Memory speeds are increased to 1,866MHz and Dell has implemented its patented Fault Resilient Memory (FRM) feature. Supporting the new Reliable Memory Technology feature of VMware's vSphere 5.5, this allows a protected memory zone to be created for the hypervisor without taking up half the system memory.

The Dell Processor Acceleration Technology (DPAT) targets high frequency trading applications that require consistent and predictable processor performance. Intel's Turbo Boost transitions can increase latency and jitter but DPAT allows you to fix the number of active cores and Turbo Boost speed to minimise these effects and provide a smoother performance.

Our review system also came with the CacheCade feature of Dell's PERC RAID controller enabled. This is designed to use SSDs to improve read operations and our lab tests show what a big difference it can make.

The R720 is very well designed and its 2U chassis offers an impressive range of feature and expansion options.

The v2 Xeons

The E5-2600 v2 Xeons see a drop to a 22nm fab allowing them to deliver more cores and better performance without increasing power consumption. Dell supplied the review system with a pair of E5-2680 v2 Xeons which are members of Intel's Advanced team of CPUs.

These new 2.8GHz Xeons are 10-core models, have 25MB of L3 cache, support 1,866MHz memory speeds and have a TDP of 115W. The first generation E5-2680 Xeon runs at 2.7GHz, has 8 cores plus a 20MB L3 cache and a higher 130W TDP.

The E5-2600 v2 family is big with the Advanced, Standard and Basic models bringing the count to seventeen. The E5-2680 v2 models in the review system will set you back around 1,500 each but you can choose lower cost models as the R720 supports the entire family.

Dell's CacheCade is quick to configure and delivers an enormous boost to read IOPS.

CacheCade performance

The review system included four 146GB SAS 2 SFF drives and a pair of 200GB Intel S3700 SATA SSDs. The CacheCade feature is accessed from the PERC's BIOS menu where you select the SSDs and create a virtual drive.

CacheCade is supported on the PERC H700 and H800 cards with 1GB of memory but Dell has still only implemented v1.0 so it will only cache read operations. You can use as many drives as you want and CacheCade creates a RAID-0 stripe from them.

Dell's iDRAC7 provides quality remote management features and can be easily upgraded to the Enterprise version.

Internal features

Dell hasn't changed anything significant internally but there's no need to as the R720 is already very well designed. The CPU sockets are accompanied by 24 DIMM slots supporting a maximum of 764GB and the price for the review system includes a healthy 128GB of 1,866MHz DDR3.

Cooling is handled by a bank of six hot-swap fans behind the drive backplane and noise levels are very low. The server also supports Dell's Fresh Air initiative for reducing data centre cooling costs and is certified to run reliably at up to 45 degrees Centigrade.

Dell scores over HP for virtualisation duties as its dual SD memory card controller is still unique. Fitting in a dedicated slot at the back of the chassis, it mirrors the primary boot media to provide hypervisor redundancy.

Expansion potential is excellent with seven PCI-e Gen 3 slots up for grabs. You have room for one 16X and three 8X full-height, full length cards and three 8X half-length, half-height cards the R720 also supports two double width GPU cards such as Nvidia's Tesla.

Remote power controls and monitoring can be accessed directly from the iDRAC7 web console.

Power and remote management

We hooked the server up to our inline meter and recorded a draw of 115W with Windows Server 2012 R2 in idle. Under maximum load from the SiSoft Sandra benchmarking app, this peaked at 322W.

In our exclusive review of the original PowerEdge R720 (web ID:639349), we measured this system drawing 120W in idle and 358W under load. The new E5-2600 v2 Xeons deliver on their power promises as the original system has a pair of 2.7GHz E5-2680 Xeons, half the memory and one less SFF hard disk.

The price we've shown includes Dell's iDRAC7 remote management controller with the Enterprise upgrade. It's well designed web interface provides a wealth of information about server operations and includes power graphing and monitoring.

You can start with the basic iDRAC7 Express card which is easily upgraded. The dedicated management port and front mounted vFlash slot for the Enterprise version are already fitted and just need a license key to be activated.

Conclusion

The PowerEdge R720 is an impressive rack server that crams an excellent range of features into its 2U chassis. Support for the new E5-2600 v2 Xeons give it even more processing density without an increase in power consumption and the CacheCade feature accelerates performance for read intensive applications.

Verdict

Dell’s PowerEdge R720 just keeps on getting better, making it well suited to running a wide range of business applications. Its E5-2600 v2 Xeons deliver more cores for less power, expansion potential is superb and its CacheCade feature can seriously boost read IOPS.

Chassis: 2U rack CPU: 2 x 2.8GHz Xeon E5-2680 v2 Memory: 128GB 1,866MHz DDR3 (max 768GB) Storage: 4 x 146GB Dell SAS 2 SFF; 2 x 200GB Intel S3700 SSD RAID: Dell PERC H710P Mini with 512MB cache/BBU/CacheCade Array support: RAID0, 1, 10, 5, 6 Expansion: 7 x PCI-e Gen3 slots Network: 4 x Gigabit Power: 2 x 750W hot-plug supplies Management: Dell iDRAC7 Enterprise with 10/100/vFlash Warranty: 3yrs on site NBD

The general purpose 2U rack server introduces IBM's pay-as-you-grow design. This allows businesses on a tight budget can start with a basic specification but its high expansion potential allows them to upgrade as required.

You can choose from a wide selection of storage and RAID options. Our review sample was fitted with an eight hot-swap SFF SAS/SATA drive cage and you can upgrade to sixteen bays when the time is right.

Usefully, the 8 Pac upgrade has a SAS expander card on its backplane. Unlike HP's latest ProLiant DL380p Gen8, you don't need a second RAID card and so won't lose a PCI-e slot.

Those with deep pockets and a need for speed can order the server with up to 32 of IBM's 1.8in SSDs. If capacity is more important, you can have six hot-swap LFF hard disk bays and then there's the CTO (configure to order) model which cuts costs further by using six cold-swap SATA bays.

The x3650 M4 has a well designed interior which offers plenty of room to expand

RAID and remote management

IBM has improved its RAID options as the base ServeRAID M5110e controller is embedded on the motherboard. This supports 6Gb/sec SAS and SATA drives but if you want more than stripes and mirrors you need an upgrade to get RAID-5 and -50 plus another if you want RAID-6 and -60 arrays.

Cache starts at 512MB with a battery pack or you can go for 512MB or 1GB of flash backed cache. As RAID is based on LSI's SAS2208 chip, you have the option to activate its CacheCade feature which optimises read activity from an SSD based cache.

The x3650 M4 showcases IBM's new IMM2 embedded management controller. The basic version is of limited use as it doesn't support web browser management or KVM-over-IP remote control.

Our system included the IMM2 Advanced upgrade which activates the dedicated Gigabit port at the back. The browser interface sees a substantial redesign over the older IMM and provides a lot more information about critical components. However, it comes up short of Dell's iDRAC7 for features and is beaten soundly by HP's new iLO4 controller.

The new IMM2 management interface includes KVM-over-IP remote control with the Advanced upgrade

IBM System Director

IBM's System Director utility is much easier to use than Dell's Management Console or HP's Insight Control software. Designed to manage all network devices, it provides discovery, software deployment, inventory, file transfer and VNC based remote control tools. The Active Energy Manager plug-in talks to the IMM2 and provides power capping plus trend graphs of power consumption and system temperatures.

IBM is lagging behind in the deployment stakes as you still have to boot its servers with the ServerGuide DVD to get an OS on it. Dell did away with the need for this task three years ago and HP's Gen8 servers have the new Intelligent Provisioning feature.

IBM's System Director is easy to deploy and the Active Energy Manager plugin provides power monitoring graphs and trending

The server is virtualisation ready as its internal USB port uses a key to boot into VMware ESXi or vSphere 5. However, HP's DL380p Gen8 provides internal USB and SD card slots but Dell goes one step beyond as its PowerEdge R720 has dual internal SD card slots for hypervisor redundancy.

Memory choices

The server's 24 DIMM slots are all enabled in dual CPU configurations. The maximum 768GB is supported by 1333MHz LR (load reduced) DIMMs, standard RDIMMs are limited to 384GB and cheaper UDIMMs are good for 64GB.

The x3650 M4 is one of the first of IBM's servers to be offered with HyperCloud DIMMs (HCDIMMs). These are expensive but support up to 384GB running at 1333MHz with three DIMMs per channel and claim a 25 per cent performance boost over standard RDIMMs.

Along with six PCI-e slots, the x3650 M4 has quad Gigabit ports plus a mezzanine connector for dual 10GbE cards

Network connections are more flexible than HP as the server has four embedded Gigabit ports plus a dedicated connector for an Emulex dual-port 10GbE mezzanine card. The DL380p Gen8 has one connector which supports either quad Gigabit or dual 10GbE mezzanine cards.

The price includes a pair of 750W hot-plug supplies which will cover most workloads but you can opt for 550W or 900W modules. With Windows Server 2008 R2 Enterprise idling along, we recorded a draw of 98W and with SiSoft Sandra putting the processors under maximum load this peaked at only 222W.

Internal cooling is handled by a bank of four hot-plug fans in front of the processors. With a DL380p Gen8 sitting next to the x3650 M4 on the lab bench, we found noise levels to be very low with nothing between the two.

Overall

If you want the best remote server management features on the planet then check out HP's DL380p Gen8 and its new iLO4 chip. However, these do come at a premium price making IBM's x3650 M4 a good choice for SMBs that want a more affordable 2U rack server that has the capacity to grow with them.

Verdict

SMBs and enterprises looking for a general purpose 2U rack server or a solid platform to run their critical apps on will find IBM’s new x3650 M4 a worthy candidate. Management features aren’t as sophisticated as those found in HP’s new ProLiants but it is good value and is scalable.

Chassis: 2U rack

CPU: 2 x 2.3GHz Xeon E5-2630

Memory: 16GB 1.35V DDR3 RDIMM memory expandable to 256GB (768GB with LRDIMMs)

Storage: 2 x 300GB IBM 10k SAS 2 hot-swap hard disks (max. 16)

RAID: IBM ServeRAID-M5110e with 512MB cache and BBU

Array support: RAID0, 1, 10

Expansion: 6 x PCI-e Gen3

Network: 4 x embedded Gigabit

Power: 2 x 750W hot-plug supplies

Management: IMM2 Advanced

Software: IBM Systems Director 6.3

Warranty: 3yrs on-site NBD

Dell's C-Series of PowerEdge servers are used in large datacentres where power consumption and cooling are higher priorities pure performance.

Aimed at markets such as cloud hosting providers, the C6220 offers four completely independent, hot-swappable server nodes. The chassis also accepts a pair of dual-height nodes which have two PCI-e expansion slots.

The range has always had a high processing density and the latest PowerEdge C6220 crams four Xeon E5-2600 servers into a low profile 2U form factor.

The four nodes are accessed at the rear where each one provides dual Gigabit, IPMI, pairs of USB2, VGA and serial ports. They're easy enough to remove by releasing a small locking tab and sliding them out rearwards.

The four hot-swap server nodes are completely independent and have their own set of ports

Improved thermal design

The older PowerEdge C6100 could only support CPUs with a maximum TDP of 95W. With the C6220, Dell has rejigged the node's thermal design so the system can cope with TDPs up to 135W.

This means the entire range of E5-2600 Xeons is supported right up to the high performance 2.9GHz, 8-core E5-2690. The review system packs in a good processing density for the price as each node was supplied with dual 2.6GHz E5-2670 Xeons.

Memory options are more limited as each node has 16 DIMM sockets as opposed to the 24 sockets found in standard Xeon E5 servers. It also currently supports 16GB 1600MHz RDIMMs so maximum capacity per node is 256GB. LR (load reduced) DIMMs will not be supported but Dell aims to have 32GB RDIMM certified soon.

The node's IPMI controller offers a comparatively basic set of remote monitoring features

Expansion and storage potential

Each node has a single X16 PCI-e Gen3 slot supporting a half-height, half-length card. If you want more, the dual height nodes provide two X16 PCI-e Gen3 slots and have room for full-length cards.

Basic storage options are provided by the motherboard's Intel C602J chipset which offers six 3Gb/sec SATA II ports and supports mirrors, stripes and RAID-5 arrays. Dell has an optional LSI 9265-8i PCI-e card which adds SAS to the mix.

Each node has a mezzanine card connector as well. Dell offers an LSI 2008 SAS card plus Gigabit and 10-Gigabit options but bear in mind this slot is only active in dual CPU configurations.

The chassis has room up front for up to twelve hot-swap LFF hard disks. The backplane is split vertically into four groups so each node has three dedicated drive bays assigned to it.

Dell offers a chassis version with 24 SFF bays so each node can have six drives. A new feature of this model is an expander backplane option that allows any number drives to be assigned to specific nodes regardless of their physical location.

No power usage monitoring but remote management provides power controls for each node

Low power usage

The chassis has room for a pair of hot-plug power supplies and our system had two 1200W modules. Dell also offers a 1400W option for heavier workloads.

To test power consumption, we hooked the C6220 up to our inline power meter. With each node loaded with Windows Server 2008 R2 and idling along we saw first, second, third and fourth nodes draw a total of 135W, 200W, 290W and 350W, respectively.

Using the Sisoft Sandra benchmarking app to push the CPUs to maximum load on each node we saw power peak at 364W, 671W, 945W and 1242W. Dell estimates the C6220 will save you around 100W of power over four servers.

In our review Boston Value Series 380 G8 which has the same CPUs and memory configuration, we saw it pull 162W in idle and 349W under extreme load. An average peak of 310W per node for the C6220 shows power savings can be made.

The C6220 is reasonably well constructed but not to the same level as Dell's standard PowerEdge servers. All cooling is handled by four fans behind the disk backplane but these are not hot-swappable and don't have integral power connectors.

Node remote control and virtual media services come as standard and not as optional upgrades

Remote management

Remote management features are basic as the C6220 doesn't support Dell's iDRAC7 controllers. The nodes have an embedded IMPI controller which links up with the dedicated Fast Ethernet port at the rear.

Its web interface provides a simple set of management and monitoring tools. You can view temperatures, voltages plus fan speeds and link these values with thresholds. Alerts can be issued if they are exceeded or you can set the node to be automatically powered down or rebooted.

No power usage monitoring tools are provided but you can remotely switch the node on and off, reset it or gracefully power it down. Whereas the iDRAC7 requires an upgrade to get KVM-over-IP remote control and virtual media services, the C6220 provides these as standard.

Overall

The C6220 doesn't save you much cash over a quartet of standard 1U rack servers as we rustled up four 1U PowerEdge C620s from Dell's web site with the same specification for a similar price. Build quality and features of the C6220 aren't as good either so it doesn't meet the objective it sets out to achieve.

Verdict

The C6220 offers a high density rack with plenty of E5-2600 Xeon muscle under the covers. It comes with impressive power to processing ratio, but is not much cheaper than a set of standard 1U rack servers. Remote management features and build quality are also basic, which makes it hard to recommend.

Chassis: 2U rack Power: 2 x 1200W hot-plug supplies Storage: 12 x hot-swap LFF SATA II drive bays Four hot-swap nodes each with the following: CPU: 2 x 2.6GHz Xeon E5-2670 Memory: 32GB 1600MHz DDR3 RDIMM expandable to 256GB Storage: 1 x 500GB Dell SATA II hard disk in hot-swap carrier (max 3 per node) Storage controller: Intel C600 RAID: RAID0, 1, 10, 5 Expansion: 1 x PCI-e Gen 3 slot, mezzanine slot Network: 2 x Gigabit, dedicated 10/100 IPMI port Management: Embedded Dell IPMI controller Warranty: 3yrs on-site NBD

The PS6100XS is one of six new PS6100 appliances introduced by Dell, which use automatic tiered storage and a hybrid array of SSDs and SAS hard disks. It's specifically designed to deliver the high IOPS needed by mission critical databases, VDI and big Exchange servers.

An important change to the entire EqualLogic family is the wholesale move to Dell manufactured hardware. This brings a significant drop in prices and comparing the PS6100XS with the Nimble Storage CS240 (65,540 Ex VAT) puts this in perspective.

Dell has redesigned its controllers which support the new cache-to-flash feature. In the event of a complete failure, a capacitor provides enough juice for the controllers to write their cache contents to flash memory.

The primary controller ports are all grouped under a virtual IP address and load balanced. As you add more appliances to a group, their network ports will join the virtual IP address so increasing available bandwidth.

The controllers in the PS6100XS each have four Gigabit ports and support Dell's new vertical port failover feature

Data tiering

The controllers don't run in the traditional active/standby mode as they use Dell's new vertical port failover. If a port on the primary controller fails, it will use the corresponding port in the secondary controller.

The appliance comes with all drives preconfigured in an accelerated RAID-6 array. Data tiering is fully automated where hot data consisting of 15MB pages is placed on the SSDs for faster access.

Expansion potential is excellent as up to sixteen appliances can be placed in a group and presented to the network as a logical storage pool. Although each appliance looks after its own RAID arrays, the storage on all members is made available as a single entity.

Volumes created within this space are presented as iSCSI targets but the volume data is spread across up to three group members which are chosen at random. This makes it a cinch to add more appliances to a group and they can be removed just as easily as their data is evacuated to other group members.

Volume creation and watermarks

Dell's Remote Setup Wizard searches for new arrays and the Group Manager console provides easy access to all functions. Management security has also been improved as Active Directory (AD) is now supported for role based access.

For volume creation you choose a storage pool, decide on a volume size and add access restrictions based on CHAP authentication, IP address or initiator name. Thin provisioning is configured during volume creation and you decide how much physical space to start with.

Three watermarks are applied where the minimum size is 10 per cent. When the volume is 95 per cent full, the system throttles back I/O performance to allow administrators time to increase the volume size. Once you reach the critical watermark the volume can be placed offline.

Snapshots schedules are applied to volumes or groups and thin provisioning uses watermarks to manage volume usage

Dell's SAN HeadQuarters

Up to 512 snapshots per volume and a total of 10,000 per group are supported. You can also group selected volumes together and apply one snapshot schedule to them all.

Replication stores snapshots of selected volumes in separate groups called replication partners. You can choose one-way replication to another group, bi-directional replication between groups or designate one group as a central location to which multiple groups can be replicated.

Dell's SAN HeadQuarters comes as standard and is designed to centrally manage multiple groups. Detailed capacity reports show the total space for each group and how it's split up into volumes and snapshots.

Graphs show total group storage, I/O capacity, network throughput and IOPS along with available headroom. New features in this version include support for multi-site management, group diagnostics and a RAID evaluation service.

Dell's SAN HQ provides detailed information about all storage related activities

IOPS galore

For performance testing we used a Dell PowerEdge R710 rack server equipped with dual 2.53GHz E5540 Xeons plus quad Gigabit ports and ran Windows Server 2008 R2. We logged the server's iSCSI initiator onto the group's virtual IP address to enable load balancing and connected it to a 100GB virtual volume.

MPIO links are deftly handled by Dell's host integration tool (HIT) as you choose how many sessions you want, log on to the portal and leave it to do the rest. The recommended MPIO setup is to enable jumbo frames, flow control and least queue depth.

The appliance's data tiering can be seen in action here as our Iometer test file has been automatically moved on to the SSDs

Configuring Iometer with four disk workers and 64KB sequential read transfer requests, we saw an quick raw read throughput of 472MB/sec. To test maximum IOPS we changed the Iometer block size to 4KB and saw a high reading of over 40,000 IOPS.

For general SQL database performance testing, we used 256 outstanding I/Os, a 16KB block size, 66 per cent read, 34 per cent write and a 100 per cent random distribution and saw 8,550 IOPS. From the SAN HQ console we could see that activity for each test was initially spread across all drives but after a few minutes, Dell's data tiering had moved all the test data to the SSDs.

Overall

The PS6100XS clearly delivers on Dell's performance promises and also scores highly for ease of deployment and expansion potential. The revamped manufacturing processes have had a big impact on costs and this is made all the more impressive as the price includes all features as standard.

Verdict

Dell’s EqualLogic PS6100XS stands out from the IP SAN crowd thanks to its ease of use, expansion potential and excellent performance. Our tests show the appliance can deliver the high IOPS demanded by heavy duty applications and Dell’s pricing structure means we have no hesitation in recommending the PS6100XS.

Chassis: 2U rack Power: 2 x 700W hot-plug supplies Storage: 7 x 400GB SFF SSDs, 17 x 600GB SFF SAS 2 hard disks (all hot-swap) Two hot-plug controllers each with the following: Cache: 4GB with cache to flash capacitor RAID: 5, 6, 10, 50, 6 Accelerated Network: 4 x Gigabit data ports, 10/100 management port Management: CLI, web browser, Dell SAN HeadQuarters

Dell has launched a new rack-mounted workstation that can support four concurrent graphics sessions.

The product was originally launched last year, but has only just received certification from Citrix to run a multi-GPU pass-through configuration.

This means that four XenDesktop sessions can run from one machine for four users performing graphics-heavy tasks.

"The Dell Precision R5500 offers the highest user density available on a XenServer certified platform for XenDesktop HDX 3D Pro, with four concurrent users per host," said Derek Thorslund, director of product management for HDX technology at Citrix.

"This makes it a very attractive option for customers looking to deliver high-end 3D professional graphics while keeping their intellectual property safe in the datacentre."

The Precision R5500 features a dual-socket Xeon processor supporting up to 192GB memory, as well as six 2.5-inch SAS HDDs or five SATA HDDs. It also boasts four Nvidia Quadro 2000 GPU adapters and runs Citrix XenServer and HDX 3D Pro remote screen technology.

Citrix XenServer with HDX 3D Pro provides the virtualised workstation environment with a dedicated graphics card per user or virtual machine.

This gives the user full access to the graphics card in a virtualised environment. Users can also remotely access the unit via HDX 3D Pro from thin clients, desktops, laptops and mobile devices using the Citrix Receiver app.

The unit is designed for users running 3D workloads. This unit is aimed particularly at organisations in finance, science, engineering, medical, and software development looking to host many users from one workstation.

UK Pricing for the R5500 starts at 1,899.

Dell consistently beats its rivals to market with new server technology, so it's no surprise the PowerEdge R820 sports Intel's latest Xeon E5-4600 processors.

When Intel launched its Xeon E5-2600 CPUs in March, their vital statistics were known well in advance, but not so with the E5-4600 family. This group consists of eight models ranging in speeds from 2GHz up to 2.9GHz with 4, 6 or 8 cores and 10MB, 12MB, 15MB, 16MB or 20MB L3 caches.

The R820 was equipped with a quartet of 6-core 2.2GHz E5-4607 Xeons which had a 12MB L3 cache, a 6.4GT/sec QPI and a low 95W TDP. These models support memory speeds up to 1066MHz and Intel's HyperThreading, but not Turbo Boost.

The E5-4600 family is designed for scalable dual and glue-less quad-socket platforms. As with the E5-2600s, they have two QPI links per socket supporting speeds up to 8GT/sec and in a 4P system will connect with adjacent sockets in a ring architecture.

In a 4P configuration the T820 uses a mezzanine card which sits on top of the main motherboard

Memory and storage aplenty

A key feature is memory capacity. The 4P system supports 48 DIMM sockets and can scale up to 1.5TB using 32GB LR-DIMMs. You may think this puts it in direct competition with the Xeon E7 but Intel has positioned the E5-4600 as a better choice where a balance of cost, performance, power consumption and memory capacity is required.

The R820 uses a similar chassis to the PowerEdge R720 and so has a good range of storage options. It supports 16 SFF hard disks and there's also an option with eight standard SSF bays and four hot-plug PCI-e Express Flash SSDs.

RAID options start with the basic PERC H310 entry-level adapter but review price includes the top dog PERC H710P PCI-e card which supports RAID-6 and has a 1GB NVRAM cache so doesn't need a battery backup. It uses an LSI chip and you can enable its optional CacheCade feature which optimises read activity from an SSD based cache.

Dell's new iDRAC7 management controller provides comprehensive power monitoring tools

Configuration

The server is well built, but it's the specification which is impressive. Squeezing quad CPUs and 48 DIMM sockets into a 2U chassis is no mean feat and Dell has achieved this with a mezzanine card.

In the base 2P system, the main motherboard in the chassis base has two CPUs and 24 DIMMs sockets. The 4P model has a mezzanine card sitting on top which provides the other two CPUs and remaining DIMM sockets.

Everything works well and the upper board can be removed easily by lifting it up using its clamping bracket. The R820 can be ordered in a 2P configuration and the 4P upgrade includes the mezzanine card. The price of the review system also includes a generous 96GB of memory.

Remote control requires an iDRAC7 Enterprise upgrade but you only need to apply a license key

Low power consumption

All chassis cooling is handled by a bank of six hot-plug cooling fans at the front. We found the R820 to be remarkably quiet during testing. The PowerEdge also supports Dell's Fresh Air initiative, which allows data centres to cut cooling costs by running at higher temperatures.

The R820 has been certified by Dell to run reliably at up to 45 degrees Centigrade for up to 1 per cent of its annual operational time. Furthermore, if data centre air conditioning fails, it can run for longer periods in uncooled environments.

The server includes a pair of hot-plug 1100W power supplies and Dell also offers 750W and 1100W DC versions. The R820 hits the spot for power consumption as our tests showed it to be undemanding.

With Windows Server 2008 R2 in idle, we measured a draw of 164W. With the SiSoft Sandra benchmarking app punishing all 48 logical cores, this only rose to 365W. Stacking this up against a Xeon E7 system puts it firmly into perspective. Fujitsu's RX600 S6 and its quartet of 10-core 2GHz E7-4850 Xeons pulling 427W in idle and a hefty 760W under load.

Remote management

Dell gives you a choice of network ports as the R820 has a proprietary slot at the rear which accepts Broadcom and Intel based NIC daughtercards. The review system came with an Intel quad-port Gigabit card but you can opt for a version with dual 10GBase-T/dual Gigabit ports which includes FCoE offload on the 10GbE ports.

There's plenty of room to expand as even with the RAID card in residence, there are still six PCI-e Gen3 slots available. Dell is moving into VDI applications and offers NVIDIA Q2000 GPU cards for the R820.

Along with the iDRAC7, you can remotely monitor the server via the browser based OpenManage Server Administrator

For remote management you can start with Dell's new iDRAC7 management controller. This slick web interface easily matches HP's iLO3 when it comes to features. If you want remote control and virtual media services you need to upgrade to the Enterprise version which has been made much easier as all the hardware is already in place and just requires a software key to activate it.

Overall

Dell is vying closely with HP for the second spot in the worldwide 4P server market, according to analysts at IDC. The PowerEdge R820 undoubtedly strengthens Dell's position as it is superbly designed and is the first Xeon E5-4600 system.

Verdict

The PowerEdge R820 showcases Intel’s latest E5-4600 Xeons in a superbly built system packed with features. The modest dimensions will appeal to space constrained data centres looking a balanced combination of processing density, high memory capacity, low power consumption and value.

Chassis: 2U rack CPU: 4 x 2.2GHz Xeon E5-4607 Memory: 96GB 1.35V DDR3 expandable to 1.5TB (LR-DIMMs) Storage: 5 x 300GB Dell SAS 2 SFF 15K hot-swap hard disks (max. 16) RAID: Dell PERC H710P with 1GB NVRAM cache Array support: RAID0, 1, 10, 5, 6 Expansion: 7 x PCI-e Gen3 slots Network: 4 x Gigabit Power: 2 x 1100W hot-plug supplies Management: Dell iDRAC7 Enterprise with 10/100 and 8GB vFlash card Software: Dell Management Console Warranty: 3yrs on site NBD

The latest version of Microsoft's data platform, SQL Server, brings enhanced scalability and flexibility, the company claimed today as it released the software to manufacturing.

SQL Server 2012 will also help customers manage and analyse more data than ever before, according to Microsoft.

"Data is being generated faster than ever before, and organisations need a way to process and analyse all that data," said Ted Kummert, corporate vice president at Microsoft.

"Whatever the type or size of data, SQL Server 2012 delivers the platform and familiar tools to manage data, generate actionable insights and help drive business impact."

With SQL Server 2012 and integrated business intelligence tools, we're processing massive volumes of data queries in near-real time.

Those tools include PowerPivot for Excel 2010 and Power View, and Microsoft said that by the end of June they will also include an Apache Hadoop-based service for Windows Azure.

Early users of SQL Server 2012 said that the software has lived up to its promise.

"Our business depends on delivering customers fast, detailed insight into hundreds of terabytes of social-network data," said David Mariani, vice president of engineering at influence measurement specialist Klout.

"With SQL Server 2012 and integrated business intelligence tools, we're processing massive volumes of data queries in near-real time."

"SQL Server 2012 Enterprise with AlwaysOn gives us exactly the performance we need," added Thomas Pullen, database administrator at BETONSOFT.

"We can sustain 1,500 game rounds per second that gives our players a much faster, better gaming experience."

Microsoft said that next month it would introduce new data warehousing solutions that build on SQL Server 2012.

These include a major software update and new half-rack form factors for Microsoft Parallel Data Warehouse appliances, as well as availability of SQL Server Fast Track Data Warehouse reference architectures for SQL Server 2012.

Partners have already begun announcing support for SQL Server 2012. One of the first was EMC, which said it had optimised its storage hardware and software to work with the new Microsoft platform.

It claimed that combining SQL Server 2012 with its unified storage, storage tiering and flash caching capabilities could quadruple database performance and yield 80 per cent faster server provisioning.

Iceotope today announced a new technique of cooling servers, which it claims can make fans redundant and cooling costs non-existent.

The launch happened at this year's CeBIT conference in Hanover, Germany, with the company also saying the solution cuts 75 per cent of the mechanical costs associated with cooling servers.

Cost savings on the maintaining side might be plentiful, but the set-up price is no drop in the ocean

The product uses 3M's Novec liquid as a coolant, which Peter Hopton, founder and chief technology officer (CTO) of Iceotope, called the "environmentally friendly little brother" of Fluoroinert 3M's previous coolant incarnation.

A water jacket containing Novec wraps around components and travels through the servers in fast currents at a rate of centimetres per second. With its low surface tension, the coolant "gets into all the cracks" to absorb the heat from servers and take it away to heat exchanges, where the liquid can either be cooled or used to heat other buildings and water within the complex.

The technology can not be incorporated into existing servers, however, and businesses would have to buy new servers or modules as Iceotope refers to them to install the technology.

Cost savings on the maintaining side might be plentiful, but the set-up price is no drop in the ocean.

Customers can buy cabinets, featuring six server modules which look like blades as well as the pump and heat exchanges, for 19,995. However, the cabinets can house up to 48 modules, which start at 3,995 for fully configured servers containing two six core Xeon E5 processors, 64GB RAM, 40Gb Infiniband and SSD storage. Users are able to choose between AMD and Intel processors.

Hopton told IT Pro the technology could soon roll-out across the data centre, with products for switches, routers and GPUs in the pipeline. He defended the upfront prices though, saying the overall benefits would save a lot more money in the long term.

"No chiller equipment is needed so when you are kitting out a new data centre, you don't need to buy it," he said.

"It is not massively expensive and if you need new servers it is cost effective when you start to look at the money saved on energy consumption."

Although he couldn't confirm any UK customers yet, Hopton did say it was being implemented within UK data centres and the test company would be revealed soon.

With the launch of its twelfth generation of PowerEdge servers, Dell takes an early lead over HP and IBM by being the first to deliver a production server with Intel's new Xeon E5-2600 processors. In this exclusive review we look at the new PowerEdge R720, which sees a complete redesign over the R710.

Storage potential sees huge improvements, as this 2U chassis has room for up eight hot-swap LFF or sixteen SFF hard disks twice that of the R710. Dell also uniquely offers a version with four standard SSF bays and four hot-plug PCI-e Express Flash SSDs.

Dell also offers the R720xd (extreme disk') version that uses a different chassis. Available as a separate order, it has been modified to provide 24 hot-swap SFF bays at the front and two more at the back.

The system on review came equipped with a pair of 2.7GHz Intel Xeon E5-2680 processors. These use the new 32nm Sandy-Bridge-EP' (efficient performance) architecture that sees inter-socket QPI links doubled and their speeds boosted to 8GT/s (gigatransfers per second).

The E5-2680 is near the top of this group of 8-core server CPUs, with its large 20MB L3 cache and support for Intel's new Turbo Boost 2.0. The latter allows cores to be speeded up briefly beyond their TDP rating. During idle periods, the system accumulates a thermal budget and in times of increased activity, this is used to boost core performance for up to 25 seconds.

The server's excellent internal design has allowed Dell to pack in an impressive range of features.

The E5-2600 Xeons have four memory channels per socket and the R720 has the full 24 DIMM slots onboard. Memory can be boosted up to a massive 768GB in dual-CPU systems using 32GB RDIMMs, or the new LR-DIMM (Load Reduced) modules.

Dell's optional dual SD memory card controller has been redesigned and fits into a dedicated slot at the back of the chassis. A mirrored mode provides hypervisor redundancy by automatically keeping an onboard copy of the primary boot media in case it fails.

The new system setup menu provides direct access to the server, the iDRAC7 and other devices.

Internal design is exemplary, and Dell has avoided wasted space by arranging the CPU and DIMM sockets in a neat row across the centre of the chassis. These are supported by a bank of six hot-plug cooling fans in front and we found noise levels to be extremely low.

Dell's Fresh Air initiative also allows data centre environments to cut cooling costs by running at higher temperatures. The R720 has been certified by Dell to run reliably at up to 45C.

Power choices are good too, with our review system supplied with a pair of platinum-rated 750W hot-plug supplies. Dell also offers 495W models for lighter duties, or higher-power 1100W supplies.

The E5-2680 Xeons in the review system have a comparatively high 130W TDP but power consumption isn't excessive. With Windows Server 2008 R2 in idle, we measured a draw of 120W and this only rose to 358W under maximum load from the SiSoft Sandra benchmarking app.

Remote power monitoring and management features have been improved significantly.

Expansion options are extensive, with seven PCI-e Gen 3 slots, with room for one x16 and three x8 full-height, full length cards, plus three x8 half-length, half-height cards. The PowerEdge R720 is also suitable for graphics intensive applications, since it supports two double-width GPU cards, including NVIDIA's Tesla.

The Sandy Bridge-EP design has the IOH (I/O Hub) moved on-chip and provides forty PCI-e lanes per socket. This new architecture means both CPUs sockets in the R720 must be populated to use all seven slots and, with one CPU, there are three slots to play with.

Network choices are extensive and a proprietary slot at the rear accepts Dell's new Broadcom and Intel based NIC daughtercards. We had the standard Intel quad-Gigabit card, but Dell also offers dual 10GBase-T/dual-Gigabit versions, which include FCoE offload on the 10GbE ports.

The review system had the quad Gigabit module but Dell also offers dual 10GbE/Gigabit versions as well.

The 10GbE modules are designed to handle the extra heat generated as they have an onboard copper heatsink. A small radiator matrix extends from the back, but it doesn't impede the PCI-e expansion slot above it.

RAID options start with the embedded PERC S110 controller. This entry-level option links up with the motherboard's four-port SATA II connector, and supports stripes, mirrors and RAID-5 arrays in Windows.

Our review system also came with the new PERC H710P Mini card. This snaps into a dedicated slot, provides 512MB of cache plus integral battery backup pack, and supports up to RAID-6 for 6Gbit/s SAS and SATA drives.

The PowerEdge R720 is also the first Dell server to use its new iDRAC7 remote management controller with a new web interface that provides a lot of operational data and improved power monitoring features.

The base iDRAC7 Express doesn't offer remote control or virtual media services, but upgrading to the Enterprise version has been simplified. With iDRAC6 you needed to buy a new snap-in upgrade module, but the dedicated management port and vFlash slot are already onboard, and just need a license key to be activated.

Dell's new iDRAC7 present a freshly designed interface that's the match of HP's iLO3.

When we reviewed IBM's System x3100 M4, we liked the new BoMC (bootable media creator) service portal, as it makes light work of system and driver updates. Dell answer is RepoMan (Repository Manager), which sounds a lot cooler but essentially does the same thing.

Not only is Dell the first blue chip to bring a Xeon E5-2600 production server to market, but it also lays down a big challenge to the competition. The PowerEdge R720 packs an unbeatable range of features into its 2U chassis and does so at a very competitive price.

Verdict

As a general-purpose 2U rack server, the PowerEdge R720 looks capable of running any application that can be thrown at it. Its massive memory capacity and hypervisor redundancy make it ideal for virtualisation, while the storage capacity, expansion potential and remote management are excellent. Plus, of course, it supports Intel’s very latest Xeon E5-2600 processors.

Chassis: 2U rack

CPU: 2x 2.7GHz Xeon E5-2680

Memory: 64GB DDR3 expandable to 768GB (using RDIMM or LR-DIMM)

Storage: 5x 300GB Dell SAS 2 SFF 10K hot-swap hard disks (max. 16)

RAID: Dell PERC H710P Mini with 512MB cache/BBU

Array support: RAID0, 1, 10, 5, 6

Expansion: 7x PCI-e Gen3 slots (with dual CPUs)

Network: 4x Gigabit Ethernet

Power: 2x 750W hot-plug supplies

Management: Dell iDRAC7 Enterprise with 10/100

Software: Dell Management Console

Warranty: 3-year on-site NBD

AMD has purchased cloud micro-server manufacturer Sea Micro for $334 million (211 million) in an attempt to find its place int he ever-booming cloud market.

The chip manufacturer has always come second in the server space against major rival Intel, but hopes the new addition to its portfolio will bolster its position and get more AMD chips into cloud environments.

AMD specifically wants to help companies with web content, social networking, search and video workloads and aims to reduce complexity within the data centre, lower cost and energy consumption and increase performance.

By acquiring SeaMicro, we are accelerating AMD's transformation into an agile, disruptive innovator.

"By acquiring SeaMicro, we are accelerating AMD's transformation into an agile, disruptive innovator capable of staking a data centre leadership position," said Rory Read, president and chief executive (CEO) of AMD.

"The unmatched combination of AMD's processing capabilities, SeaMicro's system and fabric technology, and our ambidextrous technology approach uniquely positions AMD with a compelling, differentiated position to attack the fastest growing section of the server market."

SeaMicro claims its servers use only 25 percent of the power of traditional servers, whilst only taking up one sixth of the room in a data centre. It also touts 12 times the bandwidth per core.

Andrew Feldman, CEO of SeaMicro CEO, added: "SeaMicro was founded to dramatically reduce the power consumed by servers, while increasing compute density and bandwidth."

"By becoming a part of AMD, we will have access to new markets, resources, technology, and scale that will provide us with the opportunity to work tightly with our OEM partners as we fundamentally change the server market."

Feldman will become general manager a new Data Centre Server Solutions company at AMD to incorporate his firm's technology.

In sunny Las Vegas this week, HP has been hosting its global partner conference. It marks the first time the company has brought all its partners from across the world together as HP looks to continue its comeback from the chaos of 2011.

The one announcement that stood out was the introduction of the ProLiant Gen8 Servers. As more of the features were spelled out to the gathered throngs, techies began twitching in their seats in excitement. Amongst the titillating additions were GPS to show you where a troublesome server is in the data centre, improved Sea of Sensors technology to tell you the best place for a new server to be installed and even smart socket technology to automate putting in a processor without bending any of the pins.

If you are working as an IT admin, make sure you get new skills and make yourself indispensable.

Automation is the buzz word amongst datacentre providers right now. The more manual processes you can remove, the less room for human error and the higher the likelihood of lower unplanned downtime.

This all sounds great to me, but it is the other argument vendors use when selling it to IT pros that doesn't sit right with me.

HP's Mark Potter claimed in a press conference the lessening of manual processes would allow more time for IT admins to spend on innovation, coming up with great ways to run their IT department, rather than keeping the lights on.

He isn't the first. I have heard what seems like hundreds of executives say the same thing about automation, but it is just not the case.

No matter what George Osborne comes out with next, we are still in very tough economic times. I saw more people get made redundant in 2011 than I did back in 2009 and many businesses are struggling to keep it together.

The fact is, if you have IT admins who spend their time keeping manual processes ticking over and they can be replaced by a one off payment to a technology company, you could save a lot of cash by getting rid of that employee.

It may sound harsh, but it is true. You are not going to ask them to start gazing into the future for more weird, wonderful and most likely costly ways of doing IT in the datacentre, you will give them their pink slip and write the cheque to the vendor.

If cash was readily available, then yes, maybe you wouldn't be worrying so much about making pension payments or covering holiday pay for your employee, but in an environment when the real buzzword is cuts' businesses will replace man with machine.

Yes, it will make HP and its rivals cash but, if you are working as an IT admin, make sure you get new skills and make yourself indispensable to the IT department. Don't fall victim to the misleading tales of innovation at the end of the rainbow, protect your job.

The European Commission today unveiled plans to make Europe a leading light in high-performance computing (HPC).

The EC said there had been a "relative decline in HPC use and capabilities," but it hopes that will be reversed with a doubling of investment in supercomputing.

High Performance Computing is a crucial enabler for European industry and for more jobs in Europe.

The EU plans to increase investment in the industry from 630 million to 1.2 billion and run machines that can perform 1,000,000,000,000,000,000 tasks per second, otherwise known as exascale computing.

"High Performance Computing is a crucial enabler for European industry and for more jobs in Europe," said Neelie Kroes, the EC's vice president responsible for the Digital Agenda.

"It's investments like HPC that deliver innovations improving daily life. We've got to invest smartly in this field because we cannot afford to leave it to our competitors."

The EC wants to see greater use of HPC systems and services amongst industry players and SMEs.

HPC has traditionally been reserved for larger enterprises, due to the significant cost. Supercomputers can cost as much as 100 million or more just to build.

Major manufacturers do lease out usage of their HPC machines, however. Indeed, Intel told IT Pro last year that supercomputing was going to become much more mainstream in the coming years.

The EC said it would like to see "centres of excellence" established for software in scientific fields like energy, life-sciences and climate.

Andy Grant, an HPC expert at IBM, welcomed the announcement.

"It's not just good for us but good for the European economy," he told IT Pro.

"It's good that the EU is recognising importance in HPC. It's something the UK Government has been doing as well."

As for whether HPC capabilities could reach SMEs, Grant claimed there has been a lot of effort to make HPC more SME friendly.

"We can't say for certain the barrier for entry will come down but I would advise SMEs go to cloud providers and see if they can dip their feet in HPC," he added.

More intelligent server technology will create more jobs, rather than lose IT workers their jobs, HP claimed today.

At the company's launch of its new ProLiant Gen8 server range, the focus was on automation - removing manual tasks from the datacentre and taking away complex systems sucking up most of IT administrator's working hours.

However, with less need for administration in the datacentre, some believe there could be less need for IT admins themselves.

IT admins are celebrating. They are celebrating that they don't have to spend all their time keeping their data centre running, they have time to innovate.

During a panel discussion at the launch, Richard Fichera, vice president and senior research analyst at Forrester, alluded to the issue.

"Most businesses spend much more money on planned downtime than unplanned downtime... [but] unplanned downtime is a constant headache," he said.

"[However], don't make servers too smart as people like me may well be out of a job, as well as the IT people you sell to. You have to be careful of that."

When IT Pro asked Dave Donatelli, executive vice president and general manager of enterprise servers, storage and networking (ESSN) at HP, whether this increased automation would put IT admins out of work, he claimed it would actually create jobs.

"When it comes to IT budgets, 70 per cent of what they spend is running legacy applications," he said. "If you look at what businesses want to do, they don't want to spend 70 per cent of their IT dollars getting no new innovation, they want to spend that on innovation," Donatelli said.

"The biggest frustration facing IT [departments] today... is [they] aren't moving fast enough with [technology] innovations. So really what we have set out to do, be it with these servers or with converged infrastructure, is to allow them to spend more on innovation and less just keeping what they already have running.

"I think this is job positive and job creating, particularly if you look at the technology coming out, more jobs for innovating."

Mark Potter, senior vice president and general manager of industry standard servers (ISS) at HP, backed up his colleague and claimed IT admins were not concerned but excited about the technology.

"IT admins are celebrating," he told IT Pro. "They are celebrating that they don't have to spend all their time keeping their datacentre running, they have time to innovate. They applaud this [automation]."

For more information on the new ProLiant Gen8 server range launched by HP this week, click here.

HP today announced its latest range of ProLiant servers, hoping to cut out manual processes and free up time in the datacentre, leaving space for more innovation.

The ProLiant Gen8 portfolio was born after two years of research and development, costing the company $300 million (191 million), titled 'Project Voyager.'

The research began because HP realised servers needed to "change course" to keep up with "the era of the cloud," claimed Mark Potter, vice president and general manager of Industry Standard Servers (ISS) at HP.

We need to free up admins, take advantage of capacity and remove costly unplanned downtime.

"Manual processes cost so much time that could be used to innovate," he said, during a keynote speech at HP's first global partner conference in Las Vegas.

"They don't allow the customer to harness the stranded capacity in a datacentre and address growing energy needs. We need to free up admins, take advantage of capacity and remove costly unplanned downtime."

As a result, the main focus of the ProLiant Gen8 servers is automation. An update to HP's Sea of Sensors technology will enable admins to see the best place to put a server, considering power, cooling and redundancy circuit issues through one tool. Potter claimed the server can then be plugged in and ready to go straight away.

It also automatically provisions power to certain areas, making sure nothing is wasted, and agentless software continues to cut the need for manual processes.

Potter said these examples - which fall under the umbrella of HP's 'ProActive Insight' architecture - made "a brand new ground breaking architecture, to build servers with automation and intelligence to take care of itself; exactly what it should do in the cloud era."

Other new features of the range include a smart socket addition, reducing the risk of bending pins when putting in a processor - a common cause of failures in the datacentre.

Smart Drive technology combines with the automation functionality to give engineers better information on which drives are where, making sure they don't remove the wrong one - another common cause of breakdowns in datacentres.

HP listed a number of claims around performance, including 10 per cent less energy consumption, six times faster I/O for storage and three times productivity for admins, but companies will have to wait until March to get their hands on the new servers when they become generally available.

The ProLiant Gen8 family includes the ProLiant ML Tower - for remote or branch offices - ProLiant DL rack-mounted servers, and ProLiant BL blade servers and ProLiant SL scalable system servers - suited for cloud systems.

HP will be shipping servers based on ARM chips and Intel's Atom processors in the first half of this year.

Confirmation of the launch timescale came today from Dave Donatelli, executive vice president and gemeral manager of enterprise storage, servers and networking [ESSN] at HP.

The servers were born out of the 'Project Moonshot' research programme developed by HP and announced at the start of last November.

The servers will be a significant boost to British chip designer ARM, which has traditionally been overlooked in favour of Intel and, to a lesser extent, AMD x86 processors.

This will lead to all kinds of incredible advancements and change the market

Donatelli took to the stage at HP's first global partner event in Las Vegas today, telling 3,000 HP partners his company was proud of the "game changing" announcements it had made in the server space in the past 100 days.

"Instead of building out of x86, we are going to build servers out of ARM chips and Atom chips, those same chips from your smartphones," he said.

"This will lead to all kinds of incredible advancements and change the market."

Donatelli then confirmed the new servers would be shipping in "the first half of the calendar year."

The servers themselves will be aimed at datacentres looking to run more effectively, especially large enterprises and cloud providers.

Project Moonshot aims to help companies with large sprawls consolidate their data centres and, in turn, make them more energy efficient.

HP has not turned its back on x86-based servers, however, and today launched its latest range, the ProLiant Gen8 family. To read more about the new servers, which also claim to make data centres greener with a 10 per cent reduction in energy consumption, click here.

EMC today launched VFCache, a flash caching module for servers announced last year under the Project Lightning moniker.

It marks another foray away from the storage array for EMC as it looks to bring its flash expertise to the server space.

Testing showed a three times throughput improvement and a 60 per cent dip in latency, the storage giant claimed.

If that means bringing storage to the compute makes sense, that's what we'll do. And that's kind of what we're doing with VFCache.

"This is a product that extends EMC's FAST technology into the server. It is a combination of hardware and software, the hardware being a PCIe card, which is a very exciting technology because if really changes the performance landscape it's even faster than flash inside a storage array," Barry Ader, EMC's senior director of storage product marketing, told IT Pro.

"We marry the hardware with some software intelligence that sits inside of the server, the operating environment, to determine exactly what data is hot and make sure we keep the hottest data in the server, closer to the application, closer to the CPU complex."

Data held in the cache is still pushed out to the storage array for added security, Ader said.

"This is a very nice complement to the VMAX and VMX storage devices," he added.

"VFCache will work with a variety of servers that are out there. We've qualified this with IBM, Dell, HP and our Cisco UCS partners. We have qualified VFCache with the majority of server vendors that are out there."

VFCache will either be sold to IT departments directly via EMC to line up with the vendor's storage products, or through channel partners, including server and storage companies.

The main competition for EMC in this area will be Fusion-io, which has been pushing out PCIe cards to the industry a few years ago.

"The difference here is that they [Fusion-io] went after the early adopters," Ader asaid. "They sold these cards as DAS [direct-attached storage] cards rather than cache devices. So that caused a lot of problems for their customers.

"Really mainstream customers weren't able to leverage this it didn't solve their scalability, shareability protection problems."

VFCache will also be getting deduplication technology added to it in the coming months.

Thunder coming

EMC also revealed plans for an entirely separate appliance, which will essentially do the same job as VFCache but considerably quicker.

"It's for those high-end use cases. A good example might be a high-end trading application... it might not be for the mainstream."

The first Thunder devices will be trialled with early adopters in the second quarter. Wider adoption will be opened up later in the year

EMC in server space?

Ader would not be drawn into saying whether this foray into the server space would mark EMC's eventual move into actually manufacturing servers.

"As our customers are looking for different ways to manage their information, of course EMC is going to look at these different ways that we can leverage technology like flash disrupting types of enabling technology.

"There are ways for EMC to enhance value to customers by not just going into the server space. That's not the intention. The intention here is, there are different ways to create value. In some cases that might mean bringing storage to the server.

"If that means bringing storage to the compute makes sense, that's what we'll do. And that's kind of what we're doing with VFCache."

Companies such as Cisco and HP are trying to pack more into servers, whether with networking or storage capabilities.

Dell, meanwhile, recently broke off its storage relationship with EMC, hinting it would be dallying in smarter servers with potentially greater storage integration.

For anyone who's been watching Cisco's development over the past few years, they'll know the company went through a somewhat chaotic period in 2010 and 2011.

In a bid to galvanise investor confidence, the company distanced itself from the consumer sphere, streamlined operations and went back to focusing on the enterprise.

This translated into the death of Flip cameras and umi home videoconferencing, job losses and greater embracing of technologies tied to the network.

Three years ago, no one would have believed Cisco would play a role in the server market.

We caught up with Laurent Blanchard, the networking giant's vice president of enterprise, at this year's Cisco Live event to ask why IT departments should get excited about Cisco again.

Cisco is still fairly new to the server space, having only entered the market with UCS (Unified Computing and Servers) a few years back. What can you say to IT departments to convince them Cisco is a worthy contender in this industry?

I'm not here to convince you if you are not ready to be convinced but I'm going to give my point of view.

Three years ago, no one would have believed Cisco would play a role in the server market. Actually, what we did was apply our knowledge of driving a huge amount of data through something the network into a server. We redesigned a server from scratch. When you have a legacy of many years, it is very difficult to do this. But the structure of the UCS is new from the design perspective.

In two years, we have just signed the 10,000th UCS customer. Who would have though that we would do that three years ago?

We can differentiate ourselves and the market is actually accepting it very well.

What about bringing in more storage capabilities? Is that something Cisco will do in the future rather than just partnering?

When you look at what is happening, people are moving to the cloud and we've found that there are three levels of services for infrastructure that have to be provided: compute, storage and network.

We believe we have a very strong knowledge of servers, we have 25 years+ in the network and today we believe in storage we are better off partnering with very big companies such as EMC.

We work with NetApp and we work with others as well. We have building blocks together with NetApp in the same way we have with EMC. You can say one is larger than the other one but our approach in the cloud is to drive and to work in an ecosystem.

I think the industry is understanding that the previous way of looking at IT has changed. Now the three elements are extremely integrated and you need to have a huge amount of knowledge in intelligent networks, you have to understand very well how you can virtualise in the server and therefore have the storage as capacity for those two other elements.

It is normal that the industry is going in this direction. Three years ago nobody would have believed it. I think Dell is looking at it as well, like HP, like the others.

With companies coming together in clusters to provide services, what would you say to those worried about vendor lock-in, or siloed IT architectures?

This is not what people have on their minds. First of all, a CIO of a very large organisation, they have an environment with many providers, many suppliers. They may have HP servers and Brocade solution, they don't have only one provider.

The industry is understanding that the previous way of looking at IT has changed.

Our value is that we know how to get all these pieces together and working extremely well.

There was a lot of talk about innovation during the Cisco Live keynote. Can you convince IT departments that you really are innovating when so many smaller players are making big strides?

Our key differentiator is that we know how to manage huge volumes of data transfer in an intelligent manner, in a secure manner.

If you look at the cloud, in 2010 people were saying there is roughly $30 billion investment being made into the cloud. By 2014, it will be at $100 billion and 50 per cent of the compute consumed will be cloud based.

Now if you agree with this, and if you agree with the fact that the volume of data is going to explode, and if you agree the security issue is going to be important you need to have the right network. This is where you put the most intelligence.

The innovation is to continue investing into development of the network and to answer major trends like bring your own device (BYOD), the internet of things, smart grid. Smart grid is huge business. We apply our innovation into the energy market, the smart grid.

We can make video data intelligent. You can do a search on videos now, for example. So you can search for what someone said or for keywords.

Going back to cloud, how much traction have you got within service provider data centres?

We do work with the traditional service providers, like Amazon and Google.

A lot of the talk at Cisco Live has been around providing better than 'good enough' networks, rather than premium ones like Cisco claims to offer. But a lot of IT managers are happy with good enough as long as it has got a low barrier of entry. Do you worry some of the cheaper providers are going to steal business away from you in the current climate?

I think the trend from the IT and CIO/IT manager side, the big one, is the move from Capex to Opex. When you look at the cost of running an operation, most of the cost is not the initial investment, it's running it.

You can go and buy a very cheap server with no security built into it, but it is going to cost you a lot of money after that.

But then you have the business angle which is continuously pushing you to get a new environment, new devices to be supported. Actually, what you're looking at is elasticity and agility at a cost which is the lowest cost.

We're clearly looking at the total cost of ownership (TCO), the cost of running the operation, as well as doing no compromise on security.

You can go and buy a very cheap server with no security built into it, but it is going to cost you a lot of money after that. Security is an environment where we feel there is no compromise and today it is one of the top asks of the CIO.

Can you prove that your TCO is lower than others?

We have very large companies who've done this. Let's take banking firm ING, they moved into a solution using UCS for virtualisation and they've been consolidating many servers they have IBM, HP, Dell servers. They have been consolidating using Cisco technology UCS with a VMware hypervisor and EMC. The cost of what they are running is lower than what they had before.

I don't like overall studies covering this because each environment is different, but I can tell you we would not have gotten to 10,000 customers if the competition was cheaper.

I've been at HP where I sold x86 and you would sell it, sell it, sell it. But if you don't have any differentiators, you are under an additional pressure margin. This is not where we want to play.

Barracuda's flagship products have always been its Spam and Virus Firewall appliances, and the latest 900 model reviewed here is aimed firmly at the enterprise sector. It has impressive credentials too, as along with a high claimed spam detection rate, it's designed to handle between 15,000 and 30,000 email users and up to 5,000 mail domains.

Considering the price, the 900 appliance isn't that physically impressive, since this Supermicro-based 2U rack server sports a pair of 2.6GHz 6-core Opteron 2435 processors teamed up with a modest 8GB of DDR2 RAM. For storage, there are quartets of 150GB WD Enterprise and 500GB Seagate SATA II drives in hot-swap carriers.

These drives are linked to a simple Areca PCI-X RAID controller and each group is preconfigured as RAID-10 striped mirrors. The WD drives look after the OS and message log storage, while the Seagate array provides a quarantine store.

The appliance is best located in a server room, since its cooling fans are remarkably noisy. This does show its age, as Supermicro addressed overall noise levels on its entire server range some time ago and so more recent models are much quieter.

Prices start at 31,449 and for this you get the appliance and a one-year basic replacement warranty. Barracuda's Energize Updates service for anti-spam, anti-virus and other security updates adds a further 27,749 for a three-year subscription, while a three-year instant-replacement (next business day) warranty adds another 17,649. This brings the total price to a hefty 76,847.

Despite its size and performance credentials though, the 900 is just as easy to deploy as Barracuda's small business models. The process starts with a local monitor and keyboard configuration used to complete a simple setup routine that asks for an IP address, a gateway and DNS settings.

All further configuration is via the appliance's web interface, which opens with an informative status screen showing tables and graphs of hourly and daily inbound and outbound mail traffic. It's all colour-coded, so it's easy to see what mail is being allowed, blocked, tagged or quarantined.

Barracuda's web interface provides easy access to all security features and opens with a detailed graphical overview of inbound and outbound message activity.

Next, mail recipient domains and the addresses of the mail servers the appliance will be routing mail for are added. Incoming mail must also be routed to the appliance and this can be achieved with a firewall port-forwarding rule, or by changing the DNS MX record.

For inbound mail, Barracuda uses an impressive arsenal of spam detection tools which include Bayesian algorithms, real-time blacklists, reverse DNS lookups, IP reputation analysis, fingerprinting, content filtering, keyword blocking and intent analysis. These are used to derive a score for each message and determine whether suspect messages are blocked, quarantined or tagged.

The Energize Update subscription seems high, but this does also include free access to Barracuda's Cloud Protection Layer (CPL). This can reduce local traffic by applying pre-filters to incoming mail, so spam and mail-borne viruses are stopped before they even get to the local network.

You can tweak the spam detection scoring system but we found it worked very well on the default values.

Rather than use bogus or old samples of harvested spam to test detection rates, we opted to set up the appliance to scan incoming mail in a totally live environment. This is a far more realistic test of the appliance's detection techniques, since it has to handle new and emerging threats in real time.

We left the spam scores and actions at their default settings and configured our Outlook clients to move any tagged messages into a separate folder. The appliance's message log was monitored during the test and, on completion, was exported into a spreadsheet for further analysis.

We left the appliance filtering live mail for nearly three weeks, where it handled over 6,250 inbound messages 1,447 of which were spam or carried an infected payload. Only 13 spam messages were missed, which equates to an impressive detection rate of 99.1%. The Barracuda returned a very low false positive detection rate too, with only four messages incorrectly identified as spam and seven out of 22 incorrectly labeled as possible spam.

The Energize Update subscription gives you free access to Barracuda's cloud based spam filtering service and the portal can be used to manage multiple systems.

Plenty more filtering features are provided, and custom content filters can be applied to inbound and outbound mail. A list of keywords and patterns can also be created and, if detected in the body, header, subject or attachment of a message, it can be blocked, tagged, whitelisted or quarantined.

For outbound mail, the filters can also be used to look out for business sensitive keywords and the appliance can then automatically encrypt the message before it leaves the building. Further checks can be applied to attachments, with options to look out for executable files, Microsoft Office documents and PDFs.

Keyword and pattern searches can be applied to the message body, subject and header for inbound and outbound mail.

It's also possible to automatically block or quarantine messages with password-protected archived file attachments, but the pattern matching lists for credit card and social security numbers are of limited use in the UK, since they're US-specific.

Anti-virus protection can also be extended to Microsoft Exchange servers, as Barracuda offers the necessary free agent, along with an Outlook plug-in that allows users to classify messages from their Desktop and request that selected outbound mail be encrypted.

Given the high total cost for three years' use, Barracuda's Spam and Virus Firewall 900 hardly represents fantastic value, but this does need to be balanced against the facts that it is very easy to deploy, offers an impressive range of message security functions and, above all else, delivers an impeccable real-world performance.

Verdict

Barracuda’s anti-spam appliances have always impressed us with their top-notch detection rates and the 900 model is no exception. It delivered a very high score out of the box in our real-world tests, and it was easy to deploy and configure. The hardware itself could be more up to date considering the price, but its performance and features can’t be faulted.

Chassis: 2U rack

CPU: 2x 2.6GHz AMD Opteron 2435

Memory: 8GB DDR2

Storage: 4x 150GB WD Enterprise SATA II SFF, 4x 500GB Seagate Barracuda SATA II LFF

RAID controller: Areca ARC-1120 PCI-X card with 256MB cache/BBU

Array support: Two RAID-10 arrays preconfigured

Network: 2x Gigabit

Ports: 2x USB, VGA

Management: Web browser

Warranty: One-year standard replacement

Options: Three-year Energize updates, £27,749; 3-year instant replacement, £17,649 (all ex VAT)

Along with virtualisation, Dell is aiming the R815 at businesses looking to consolidate their existing servers into fewer, more energy efficient, systems. For a 2U server, the R815 packs a heck of lot into a compact chassis. The review system came equipped with four 2.1GHz 16-core Opteron 6272 processors and the listed price includes a healthy 128GB of low-voltage RDIMM memory, too.

Dell supports the entire Opteron 6200 range, which offers a real mixed bag of core counts: one 4-core, two 8-core, two 12-core and five 16-core models. The 6272 modules in the review system have a 115W rating, but other 16-core options range from the 85W 1.6GHz 6262HE to the power-hungry 140W 2.6GHz 6282SE.

Along with virtualisation, Dell is aiming the R815 at businesses looking to consolidate their existing servers into fewer, more energy efficient, systems. It certainly scores highly in the latter department, as our power tests show it to be easy on the utility supply. Our R815 had two hot-pluggable redundant 1100W power supplies and our in-line meter recorded a draw of 297W with Windows Server 2008 R2 idling.

Using the SiSoft Sandra benchmarking app, we saw peak power usage climb to around 660W. To put this in perspective, a Dell PowerEdge R810 with a pair of 2GHz Intel X6550 Xeons and the same amount of memory pulled 392W in idle and 558W under peak load. Considering the R810 server had 16 physical cores, as opposed to 64 in the R815, the energy consumption benefits of the Opteron 6200 are quite clear.

The Opteron 6200 uses the same chunky processor G34 socket as the 6100 family and is essentially a drop-in upgrade for existing systems. In reality few businesses will actually want to do this, but Dell does offer kits for those that do. What it does mean is that server vendors can take existing Opteron 6100 systems, tweak their BIOS and serve them up again with a higher core count.

The chassis for the R815 is real all-rounder too, since not only is it used for the Opteron boards, but it's also virtually identical to Dell's R810 Xeon 6500/7500 based servers.

Expansion potential is very good and the R815 has six PCI-e slots, though they're unlikely to needed the server already has four embedded Gigabit ports.The Dell R815 is very well designed, with the front panel split into two horizontal sections and the lower half providing unobstructed air flow through the chassis.

To the right are six hot-swap SFF disk bays that support SATA, 6Gbit SAS, near-line SAS and SSD drives. On the opposite side is Dell's nifty little LCD control panel. This provides a keypad for setting the network address of the iDRAC6 remote management controller, along with options to scroll through views of the server's serial number, power consumption and temperatures.

The server also has a very well designed interior that's completely tool free. The four processor sockets and their attendant banks of eight DIMM sockets are staggered to improve airflow, and the front sockets are accessed by releasing the hard disk bay and sliding the top half of the front panel forward.

Six hot-swap fans are spread across the middle of the chassis and we found the server to be very quiet during testing. Each fan can be removed individually but releasing clamps on each side allows the entire assembly to be lifted out.

RAID options are plentiful. The embedded SATA controller supports both mirrored and striped arrays, while the supplied PERC H700 RAID controller card has a 512MB cache and battery backup, and supports all the usual array suspects, including RAID-6.

Expansion potential is very good and the R815 has six PCI-e slots, though they're unlikely to needed the server already has four embedded Gigabit ports and the RAID card has its own dedicated expansion slot, too.

Rather than use a USB port, the R815 can be booted into an embedded hypervisor using Dell's optional internal SD memory card controller, which is well worth the additional 60 cost. It provides two SD card slots and automatically keeps an onboard copy of the primary boot media in case of failure. The R815 also has a Lifecycle Controller' with 1GB of NVRAM to ease operating system installation it allows the server to be booted into Dell's UEFI (Unified Extensible Firmware Interface) environment and deploy an OS without needing to use any boot media.

Two remote management options are available and the price shown here includes the full iDRAC6 Enterprise snap-in module that provides a dedicated network port and full remote access to the server. Note that the cheaper iDRAC6 Express module doesn't support full KVM-over-IP remote control or virtual media services, though.

AMD's main competition for its Opteron 6000 processors has always been Intel's Xeon 5600 and it beats them hands-down for the sheer core count. It's also a lot better value and more power efficient that current four socket Xeon 7500 systems. Intel's Xeon E5 is due to be announced at the beginning of March but for the moment, AMD has the upper hand in the server core wars and its Opteron 6200 makes Dell's PowerEdge R815 a fine choice for virtualisation and server consolidation duties.

Verdict

With its quartet of 16-core Opteron 6200s and high memory capacity, the PowerEdge R815 is a particularly good candidate for virtualisation and as a server consolidation platform. It packs a lot of useful features into its compact 2U chassis, has a high expansion potential, and is more power efficient and far better value than Xeon 7500 based systems.

Chassis: 2U rack

CPU: 4x 2.1GHz 16-core AMD Opteron 6272

Memory: 128GB DDR3 LV RDIMM expandable to 512GB

Storage: 6x 73GB Dell 15K SFF 6Gbps SAS hard disks in hot-swap carriers

RAID: Dell PERC H700 with 512MB cache and BBU

Array support: RAID 0, 1, 10, 5, 50, 6

Expansion: 6x PCI-e 2.0

Network: 4x Gigabit

Power: 2x 1100W hot-plug supplies

Management: iDRAC6 Enterprise with VFlash/SD Card slot and 10/100 port

Software: Dell Management Console

Warranty: 3-year ProSupport on-site NBD

Dell's Data Center Solutions (DCS) group is one of the company's least well known divisions, but also one of the largest and most profitable. It will build any server you want to your own specifications, providing you order a few thousand of them at a time.

DCS has learnt a few things during this time and has used its field experience to produce a commercial range of rack servers for large scale customers. The PowerEdge C-Series family consists of seven products and in this exclusive review we look at the C6100.

The C6100 is aimed at enterprise datacentres running HPC or cloud services where power, space, cooling and maintenance are high priorities.

The C6100 is aimed at enterprise datacentres running HPC or cloud services where power, space, cooling and maintenance are high priorities. The C6100 gets the ball rolling with a high rack density as it combines four independent server nodes into a low-profile 2U chassis.

From the front, the C6100 appears to be a standard storage server with twelve hot-swap 3.5in disk bays. However, the chassis' backplane has been designed to provide each server node with three dedicated bays arranged vertically into separate groups.

Dell also offers another version with 24 2.5in disk bays so each server can have a total of six drives. The price for the review system includes eight 250GB 3.5in SATA hard disks. The rack brackets on each side of the chassis incorporate power buttons for each node along with a single warning light.

Each server node is totally independent, with its own monitor, network, serial and USB ports.

Moving to the back of the chassis confirms that each server node is, indeed, totally independent and has its own monitor, network, serial and USB ports. The nodes are hot-swappable so you can remove one whilst the others are running. A node is removed by pressing on a small locking tab next to its expansion bay and sliding it out the back using the handle.

The nodes have a PCI Express X16 riser for general expansion and a mezzanine slot for Dell's optional Infiniband, RAID or 10-Gigabit cards. You may need to use one as the embedded ICH10R SATA controller doesn't have any RAID options.

If you want RAID, Dell offers an LSI 6Gbit/s SAS mezzanine card which supports mirrors and stripes or an LSI MegaRAID SAS PCI-E card which adds RAID 5 and 6 to the mix. The backplane itself supports both SATA and 6Gbit/s SAS disks. For the latter the backplane has separate four-port mini-SAS connectors for each node that are cabled directly to the cards.

The chassis includes a power throttling feature designed to protect the system and running services.

Along with sharing out the disk bays to each node, the chassis incorporates a pair of 1100W hot-plug power supplies. Power is rerouted through the same backplane as the disks so the nodes receive all storage and power services via a single edge connector.

Dell also offers 470W, 750W and 1400W supply options and the chassis includes a power throttling feature designed to protect the system and running services. If consumption exceeds the output of one supply then redundancy cannot be maintained if one supply fails.

In this event, the system will throttle back power consumption of the processors on all nodes to stay within its power budget. Once the errant supply has been replaced, it'll return the nodes to full speed. To reduce consumption further, the supplies don't have any internal fans and rely on the four main system fans behind the backplane for cooling.

In idle we saw one, two, three and four nodes draw a total of 136W, 205W, 275W and 348W. Under pressure these figures rose to 295W, 516W, 737W and 964W respectively. An average peak of 241W per node under heavy load is low, showing just how power frugal the C6100 is.

Unlike Dell's iDRAC, the C6100 controllers include KVM-over-IP remote control and virtual media services as standard and not as an optional extra

Remote management features differ from Dell's standard PowerEdge servers as the nodes in the C6100 don't support its iDRAC upgrades. The motherboards feature a basic Intelligent Platform Management Interface (IPMI) controller that can share access with the first Gigabit Ethernet port or use the dedicated port alongside it.

Apart from some rebranding, the web console is identical to that presented by Intel's, Supermicro's and Asus' IPMI controllers. It provides a simple web browser interface where you can view all motherboard sensors for critical components.

Full control over power allows the server to be remotely switched on and off, reset or gracefully powered down. It's better value than Dell's iDRAC as KVM-over-IP remote control and virtual media services are provided as standard and not as an optional upgrade.

Despite its obvious benefits, Dell is facing very little competition to the C6100. The only real alternatives are HP's SL6500 Scalable Systems and Supermicro's Twin rack servers.

The C6100 delivers an equally high processing density in a compact rack server. Considering the specification of the review system, it is good value overall value and it has the considerable weight of Dell's DCS group behind it.

So what's our verdict?

Verdict

Dell’s PowerEdge C6100 is aimed at the middle ground between customers looking for a few rack dense systems and those requiring a few thousand. Businesses building an HPC, gaming or cloud services solution will find it has good processing density for a 2U chassis and it’s low on power consumptions as well. Supermicro was here first with its Twin² servers which are better value, but its resellers won’t be able to compete with the warranty and support services offered by Dell’s mighty DCS group.

SPECIFICATIONS Chassis: 2U rack Power: 2 x 1100W hot-plug supplies Four hot-swap nodes each with the following: CPU: 2 x 2.67GHz Intel Xeon X5650 Memory: 48GB DDR3 expandable to 96GB Storage: 8 x 250GB SATA hard disks in hot-swap carriers (max three per node) Storage controller: Embedded ICH10R SATA Controller RAID: None Expansion: 1 x PCI Express slot, mezzanine slot Network: 2 x Gigabit Ethernet, dedicated IPMI port Power: 2 x 750W hot-plug supplies Management: Embedded Dell IPMI controller Warranty: four years on-site Next Business Day

Just over a year ago, Michael Dell didn't have the full confidence of his company's shareholders.

A quarter of them withheld voting for the founder's re-election, as Mr Dell received less support than any other executive re-elected to the board.

The central issue facing the chairman and CEO was over a US Securities and Exchange Commission (SEC) investigation into how Mr Dell and his firm failed to adequately disclose information to shareholders over payments from Intel. The chip maker effectively paid Dell to keep using its processors, rather than switch to AMD.

The payments from Intel made it seem as though Dell was meeting or surpassing results expectations, the SEC said. Without them, however, the company would not have done so once between May 2001 and January 2006, the regulatory body claimed.

Dell was told to pay $100 million as part of the settlement, whilst the company founder personally paid out $4 million. It was not a pretty end to a murky story.

The legal investigation was not Dell's only problem. Some were calling for fresh blood, for a chief who could take Dell into a new era of IT in the cloud and thinner clients, as the desktop market falls.

This all came just after three years of Michael Dell's return as CEO of the company he founded, having stepped down in 2004.

Like a nerdy kid hitting the gym in earnest, Dell has bulked up to become a full data centre solution company.

He has made some bold attempts to answer questions surrounding his ability to take Dell forward over the last year, however. He has re-envisioned his company, whilst keeping a firm grip on the client market, as others namely HP have chosen to move away.

Has the company-coined "new Dell" brought him back from the brink and secured his place as the leader of the business he created?

Welcome back

From the outside, it appears so. Dell has been a confident company of late.

Back in February, it posted some solid results, as its fourth quarter smashed Wall Street expectations. Many pointed to a Dell revival, as the company shifted into higher margin business, including a greater focus on servers, storage and services.

In August, the company again posted some decent figures as net income went up 63 per cent year-over-year. However, it reduced its revenue expectations for the year down to between one and five per cent. Yet it would be harsh to slam Dell for this mixed showing, given the entire market is being hit by reduced demand, added to the fact it is going through a major transitional period. We can forgive this faltering for now.

Highlighting its commitment to change, Dell has spent a penny or two. As part of its redefining of its business to focus on overall solutions, moving away from a hardware-centric model, Dell has bought a raft of companies.

The acquisition of Perot Systems for around $3.9 billion kicked off the spending spree in 2009, marking a notable moment in Dell's services commitment. Then came KACE Systems in 2010, providing hardware-based systems management appliances, followed by storage company Compellent, having missed out on 3PAR. Its latest move was to get hold of Force10 Networks to provide some prowess in the networking space. Like a nerdy kid hitting the gym in earnest, Dell has bulked up to become a full data centre solution company.

"Finally, some of the power brokers in Dell have accepted that pile them high, sell them cheap' has had its day, and the acquisitions Dell has been making are all aimed at moving it into a much more complex, full hardware stack sell, with the software coming from solid relationships, mainly with Microsoft, but also on the Linux side as well," said Clive Longbottom, founder of analyst firm Quocirca.

"I do believe he has now seen that Dell could not go on as it was. New blood has been brought in, new approaches are being brought to the fore, and Dell is looking stronger than it has done for a while."

Lingering issues

On the face of it, Mr Dell has helped create a new company ready for today's IT world. Yet significant stumbling blocks remain. Number one is the declining PC market and Dell's falling position within it.

During its inaugural Dell World conference in Austin last week, Lenovo became the number two PC seller, pushing the US firm down into third place. Given the PC industry is already in decline, the Texan firm will not want to fall from the upper echelons of the market.

Exacerbating things is Dell's performance in other client segments. In tablets, one of the causes of the PC market's contraction, Dell has failed to impress. Its lineup of Streak devices have performed poorly, with the Streak 5 model ditched earlier this year. It has plans for a global release of 10-inch consumer-focused tablets, but they have not emerged in the West yet.

Dell hasn't made much of a mark on the smartphone space either. Despite producing a quality phone with the Dell Venue Pro, its market share remains infinitesimal. According to Gartner, its not even in the top 10.

Michael Dell may have to face up to the reality it shouldn't bet on the consumer segment as much as it currently is. Cisco has proven how disastrous major gambles in the consumer sphere can be, especially if you don't have the relevant acumen or experience.

By turning his ship towards a cloudy horizon and building hardware for the entire stack, Dell is looking like a considerably more attractive proposition.

In tablets, Dell will have to create something with top-level design quality and some significant power, as well as a low-entry price point if it wants to have any chance at succeeding. Whilst it can certainly create quality hardware, offering low costs looks unlikely. For starters, it won't be able to rely on apps and software sales like Amazon can.

Furthermore, no matter how much Dell says smartphones are part of the end-to-end solution for businesses, it can't deny the devices are primarily targeted at consumers. When you think of smartphones though, you do not think of Dell.

Simply looking at the most recent results, which showed much of the positives were coming from the enterprise segment, it should be clear to Michael Dell that placing considerable emphasis on the consumer is a big risk.

Let's get positive

Yet Mr Dell has done well. By turning his ship towards a cloudy horizon and building hardware for the entire stack, Dell is looking like a considerably more attractive proposition for companies who want a full business package.

"As long as Dell retains focus and manages to keep fighting with its broadened arsenal of compute, storage, management, network hardware, brought through with professional services and being sold as a solution sell, rather than a portfolio sell, I believe that it will begin to take business away from a wounded HP, and could also chip away at IBM where some organisations see IBM as being too expensive and cumbersome to deal with," Longbottom told IT Pro.

Michael Dell's second reign as CEO won't be of Steve Jobs proportions. He will not turn Dell into a cult-inspiring, behemoth of a tech provider, but there is little doubt he has steered the business in the right direction. All he has to do now is keep firmly on course hopefully steering away from consumer clients - amidst turbulent seas.

As IT spending looks set to diminish over the next year, Dell will have to look to those markets it knows will provide it with solid income.

Many businesses don't use demanding applications. Consequently, they see little point in investing in servers using the latest multi-core Xeon and Opteron processors when most of their computing power will remain untapped.

Boston's Green Power 2200-T is designed for such businesses. It's a low cost, high-density system for running basic applications such as web serving and hosting, mail and BlackBerry services. It uses Intel's Atom processor and brings together eight, independent server nodes into a 2U chassis.

The 2200-T is based on Supermicro's Twin Cubed server platform which consists of a 2U rack chassis and four hot-swap trays each using an X7SPT-DF-D525 motherboard. The motherboards have two embedded 1.8GHz Intel Atom D525 dual-core processors and are designed to present each processor as a totally separate server.

One of the X7SPT-DF-D525 motherboards used in the Green Power 2200-T

The chassis has two hot-swap 720W 80 Gold Plus certified power supplies and both are cabled through to the hard disk backplane at the front. The motherboard interfaces are hard-wired though to a hot-plug bridge bar on the side of the tray which mates with the chassis backplane during insertion allowing it to receive power and hard disk storage from a single connector.

The chassis provides a total of 24 hot-swap SFF hard disks so each node gets three drives apiece and the price includes two 500GB disks per node. The motherboard's embedded Intel SATA controller supports mirrors, stripes and RAID-5 so storage redundancy per node is good.

A glance at the rear of each tray shows a complete set of ports for each server node comprising VGA, two Gigabit Ethernet and two USB2 ports. Further expansion isn't possible as the motherboards don't have any PCI Express slots or room for a mounting bracket on their rear panel.

A drawback of using Atom chips is their memory limitations as the motherboards only have two SO-DIMM slots for each node. The Atom's maximum supported memory is 4GB of 800MHz DDR3 and this is also the unbuffered, non-ECC variety so we wouldn't recommend using this server to run mission-critical applications.

System cooling is handled by four large fans located between the trays and the backplane. They aren't hot-swappable and each pair can only be removed when the lower trays have been released and pushed back slightly. With the Atom's staying frosty the fans don't need to shift much air and we found the server to be surprisingly quiet even when under heavy load.

The mounting brackets on the chassis have integral power buttons so you can control each server individually. They also provide a basic set of status LEDs for power and the network ports. The trays are easy to remove as you release the locking tabs on each side of them and pull the unit out from the rear with the two grab rings.

Remote management is possible since each server has an embedded Intelligent Platform Management Interface (IPMI) controller which shares access with the first Gigabit Ethernet NIC. This offers a tidy web interface providing full access to the server so you can switch it on and off, reset it and initiate a graceful shutdown.

The web interface provides sensor readouts for all critical components and you can issue SNMP traps and email alerts using thresholds for each sensor. You're going to need full remote control for this system and thankfully it comes as standard.

The remote control came in very handy for us during our power tests as we could control all eight nodes from one workstation.

It also provides virtual media services so devices on the host PC, such as optical drives, can be used by the server under control. We found the KVM-over-IP feature particularly useful as we could remotely control all eight nodes from one workstation during our power tests.

Power usage is incredibly low as each node uses less than a 40W light bulb, even when under heavy load.

With one node powered up and the four fans running we saw usage with Windows Server 2008 R2 in idle settling at 89W. With SiSoft Sandra thrashing the Atom processor, usage peaked at only 97W. We then powered up more nodes and took idle and peak measurements as we went.

In idle we saw two, four, six and eight nodes draw a total of 108W, 146W, 195W and 235W and under pressure these figures peaked at 117W, 164W, 216W and 263W. Overall, we were seeing an average power usage per server of only 29W in idle and barely 33W under load which is quite remarkable.

Businesses looking for a high-density compute node solution capable of running low-demand applications will find that the 2200-T fits the bill very nicely. Furthermore, the price for the review system equates to only 699 per server and power usage is incredibly low as each node uses less than a 40W light bulb, even when under heavy load.

So what's our verdict?

Verdict

The Green Power 2200-T finds an innovative use for Intel’s Atom processor and is well suited to running non-critical applications that don’t need a lot of horsepower. Cramming eight servers into a 2U chassis gives it a high rack density, the cost per node is very low and data centres that have maxed out their utility supply will appreciate its ridiculously low power consumption.

Chassis: 2U rack Power: 2 x 720W hot-plug supplies Eight compute nodes each with the following: CPU: 1.83 GHz Intel Atom D525 Memory: 4GB 800MHz non-ECC unbuffered DDR3 Storage: 2 x 500GB Seagate 3Gbit/s SATA SFF hard disks in hot-swap carriers (max. 3 per node) RAID: Intel embedded SATA controller Array support: RAID 0, 1, 10, 5 Network: 2 x Gigabit Ethernet Management: Embedded Supermicro IPMI controller (shares NIC1); Software: Supermicro SuperO Doctor III and IPMI View 2 Warranty: 3yrs on-site next business day