HPE has announced an expansion to its ProLiant edge portfolio, with a particular focus placed on supporting organizations operating in “harsh environments”.

The new HPE ProLiant Compute EL2000 chassis, which supports up to two HPE ProLiant Compute EL220 Gen12 servers or one EL240 Gen12 server, is designed specifically for applications in manufacturing, retail, telecommunications, and national security.

According to HPE, the new servers boast a range of performance improvements compared to prior generations, and can scale from eight up to 144 Intel Xeon 6 cores.

The EL240 Gen12 server is also available with Nvidia RTX Pro 4500 or Nvidia RTX Pro 6000 Blackwell server edition GPUs.

Durability is a key focus here, according to HPE. The new server range can operate in extreme temperatures ranging from -40 degrees Celsius to 55 degrees Celsius, as well as environments with up to 95% humidity.

They’re also designed to compensate for heavy vibration from aircraft or ground vehicles, as well as electromagnetic interference (EMI) and environmental contaminants.

During a press briefing ahead of the announcement, Krista Satterthwaite, senior vice president and GM of HPE’s compute group, said the focus on durability and ruggedness comes in direct response to changing edge requirements.

“Every edge is different, and edge is hard,” she said. “Servers at the edge need to be tough across industries like retail, manufacturing, [and] telecommunications.”

“Workloads are running at edge locations that are understaffed, constrained by space power, connectivity, and security becomes even more critical because these environments are much more exposed, which means protection has to be built in from the start.”

HPE touts telecoms, defense potential

HPE revealed that each platform is now available with an “Environmental Ruggedization” kit, enabling organizations operating in harsh conditions to compensate for “extreme stressors”.

“We're introducing an environmental ruggedized kit that enables these servers to handle extreme stressors,” Satterthwaite told assembled media.

“These new offerings satisfy customer needs for purpose built systems that deliver highly reliable performance in harsh and highly distributed locations that are designed to operate remotely with end to end visibility and control.

According to HPE, the ruggedization features adhere to a range of standards, including US national security standards, which require validated server survivability capabilities in extreme temperatures or high or low altitudes.

Satterthwaite noted that growing demands for edge capabilities in the defense industry and the public sector have played a key role in the durability focus.

“That is one of the main focus areas, which is why we have so many things that we’ve pursued in terms of certifications for more ruggedized environments,” she commented.

“So big, big demand coming from the defense space, in particular in the public sector.”

John Carter, VP for mainstream compute at HPE, echoed Satterthwaite’s comments, noting that the systems will “go anywhere” and “withstand any environment”.

“We’ve seen demand and had customer conversations from everything from space stations to battleships, to Jeeps, to anything else you can think of,” he said.

Elsewhere, the new range also adheres to telecoms network equipment standards that support 5G and radio access network (RAN) availability in unattended environments.

On the telecommunications front, Carter told assembled media this is an industry where the company is seeing surging demand for edge capabilities.

“Telcos have really been embracing the edge for quite a while,” he said. “They’ve been expanding the 5G roll-outs, now they’re starting to look at 6G in the future, and that is requiring a new level of performance and compute.”

“But we have seen numerous different environments from different customers. It just depends on the provider. And we do work with those telco providers from everything from the extreme edge all the way back to the packet core,” Carter added.

FOLLOW US ON SOCIAL MEDIA

Global server market sales hit record levels last year, according to new research from IDC, driven by skyrocketing demand for AI infrastructure.

New figures from the consultancy show organizations continue to ramp up infrastructure spending, particularly with top hyperscalers, while large service providers are racing to lead in the AI era.

As a result, the market saw a staggering 80.4% annual growth rate, recording $444.1 billion in revenue. In the last quarter of 2025 alone, revenue reached $125.3 billion – a 52.4% year-over-year increase.

There was a year-over-year increase of 52.4% in vendor revenue compared with the same quarter of 2024. Revenue generated from x86 servers increased 16.9% in the last quarter to $69.8 billion, while non-x86 servers rose by 146.4% to $55.5 billion.

Elsewhere, revenue for servers with an embedded GPU grew 59.1% year-over-year, representing more than half of the total server market revenue.

Memory price surges are a worry

The rising price of certain components, such as GPUs, DRAM, and SSDs, is causing worry. IDC noted that some enterprises have been trying to deal with the problem by securing prices in advance.

However, the impact could be set to hit harder during 2026 as demand keeps outpacing offering capacity in the near term.

“The race for AI adoption is settling the market pace with companies starving for infrastructure looking not only GPUs but also consuming more CPUs, among other components, in order to feed their needs,” said Juan Seminara, research director of Worldwide Enterprise Infrastructure Trackers.

"We are going to see more price pressures, and that may impact on market dynamics with less units but higher average selling prices going forward."

Dell among the big server market winners

Regionally, the US saw the fastest server market growth, up 72.4% in the fourth quarter of 2024, fueled by 80.1% growth in the accelerated server segment.

Canada grew by 70.7%, for the same reason. EMEA and APeJC, meanwhile, showed double digit growth with 43.6% and 27.9% respectively. PRC and Latin America showed lower growth of 17.7% and 12.8% each, while Japan declined by 4.7%.

Dell Technologies was the clear leader in the OEM market with a 10.0% revenue share, thanks to significant growth on accelerated servers. Supermicro was in second place with 9.5% revenue share for the same reason.

IEIT Systems and Lenovo pretty much tied for third, with 4.1% and 4.0% share respectively, while Hewlett Packard Enterprise finished in fifth place, with a 3.1% share.

FOLLOW US ON SOCIAL MEDIA

Global server market growth skyrocketed in the third quarter of 2025, new research shows, fueled by continued enterprise cloud and AI spending.

Figures from IDC’s Worldwide Quarterly Server Tracker show the market hit a record-breaking $112.4 billion in revenue across Q3.

This, the consultancy noted, marked yet another double-digit growth rate, with the industry recording 61% growth compared to the same period last year.

Revenue generated from x86 servers increased by 32.8% to $76.3 billion, according to IDC. However, the biggest growth rate came in Non-x86 servers, hitting $36.2 billion and marking a huge 192.7% YoY increase.

Notably, revenue for servers with an embedded GPU now accounts for more than half of broader server market revenue, with growth in this domain surging 49.4% YoY.

IDC said a key factor here lies in sharpened hyperscaler focus on AI infrastructure projects, with major providers ramping up hardware spending to accommodate rapid enterprise adoption rates.

All told, spending on this front has helped fuel broader server market growth over the last 18 months, with IDC noting that the market has doubled in size compared to 2024, reaching revenues of $314.2 billion in the first three quarters.

“IDC expects AI adoption to keep growing at an outstanding pace as major vendors continue reporting record orders and showing strong backlogs,” said Juan Seminara, research director, Worldwide Enterprise Infrastructure Trackers.

“Hyperscalers and cloud providers are still ahead with new, large deployments that require much higher compute density. Additionally, we started to see major AI based Research and Education projects that will help fuel further growth.”

US market surge continues

IDC’s tracker shows continued growth across a host of markets, with the US in particular the fastest-growing region due to heightened hyperscaler infrastructure spending.

Indeed, the US recorded a 79.1% increase compared to the third quarter of 2024. IDC added that a key contributing factor here was a 105.% growth in accelerated server spending.

Market growth in China also surged by 37.6% YoY, with the region accounting for nearly one-fifth of global quarterly revenue.

“APeJC, EMEA and Japan also showed very healthy double digit growth with 37,4%, 31.0% and 28,1% respectively,” IDC added. “While Latin America showed a low single digit growth with 4.1% increase in the quarter.”

The big boom time winners

Major vendor positions in the server market have fluctuated over the last year, IDC found. Dell Technologies holds a solid lead, boasting an 8.3% revenue share which the consultancy attributed to "outstanding growth on accelerated servers”.

Supermicro, meanwhile, holds second place with a 4% share. However, IDC noted that its position here compared to last year has slipped by 13.2%.

Elsewhere, Lenovo and IEIT Systems remain tied in third position, commanding a 3.6% and 3.7% share respectively.

MORE FROM ITPRO

• Dedicated servers are back in vogue as IT leaders scramble to meet AI, compliance requirements
• Best business servers
• How to set up a server for small businesses

Organizations are increasingly reliant on dedicated servers, new research from Liquid Web shows, and AI performance requirements are a key factor in their revival.

In a survey of more than 1,000 IT professionals across industries and company sizes, 86% said their organization is currently using dedicated servers, with government (93%), information technology (91%), and finance (90%) the most likely to do so.

The trend appears even within smaller businesses, with 68% of micro companies reporting using them - and these numbers are growing. In the past 12 months, 42% of respondents said they've migrated workloads from public cloud back to dedicated servers.

Just over one-third of respondents (34%) said their organizations increased spending on dedicated servers in 2024 - more than in any previous year.

"This upward trend in 2024 spending reflects a broader shift back to dedicated environments, with 42% of IT professionals saying they migrated workloads away from the public cloud in the past 12 months," Liquid Web said.

In terms of workloads, they're being mainly used for databases, large-scale file storage, web hosting, internal developer environments, compliance-sensitive systems, and AI model training or inference.

Just over half of respondents said they'd never use public cloud for regulated data such as healthcare or finance, with others citing core databases, performance-intensive computing, and AI workloads.

“People outside of tech often think servers are outdated or only for legacy systems. But dedicated infrastructure still plays a critical role in compliance-heavy industries, and sometimes outperforms cloud for specific workloads," said one survey respondent.

“The biggest misconception is that we don’t need them anymore because of cloud adoption,” another respondent said. “But dedicated servers still handle our most critical operations due to reliability, control, and predictable costs.”

IT leaders eye more control with dedicated servers

More than half the IT professionals (55%) said their main reason for choosing dedicated infrastructure over cloud was full control and customization. Other reasons included network performance, physical security, predictable pricing, and custom hardware configurations.

Unfortunately, some said they faced communication challenges within their organizations, with nearly a third saying they don’t feel confident explaining infrastructure choices to non-technical leadership figures.

More than half of IT professionals (53%) said they still consider dedicated servers essential, and nearly as many (45%) believe their importance will grow even more by 2030.

However, there’s still room for improvement, with a quarter of respondents asking for easier scaling, with others looking for access to modern developer tools (15%) and sustainable energy use (14%).

"Despite predictions of cloud dominance, dedicated servers remain a foundational component of enterprise infrastructure, especially in industries where performance, control, and compliance are critical," said the researchers.

"IT professionals continue to rely on them not out of nostalgia, but because they deliver predictable, customizable environments that serve mission-critical needs."

MORE FROM ITPRO

• AI security and compliance concerns are driving a private cloud boom
• Where next for private cloud networks?
• There’s a ‘cloud reset’ underway, and VMware Cloud Foundation 9.0 is a chance for Broadcom to pounce on it

The rise in AI has helped push server sales to the second highest growth rate since 2019, nearly doubling since last year.

The AI boom has driven a massive rollout in infrastructure, with research suggesting data centres will continue to grow in size and number.

That surge is reflected in IDC's worldwide quarterly server tracker, which showed the overall market topped a record $77.3 billion in revenue in the final quarter of last year, up by 91% year-over-year.

"IDC expects AI adoption to continue growing at a remarkable pace as hyperscalers, CSPs, private companies, and governments around the world are increasingly prioritizing those investments,” said Lidice Fernandez , group vice president, Worldwide Enterprise Infrastructure Trackers

Revenue was up across all types of servers. Traditional x86 servers were up 59.9% to $54.8 billion, while non-x86 servers climbed a whopping 262.1% to $22.5 billion.

GPU growth reflects AI boom

Servers with an embedded GPU were up 192.6% for the fourth quarter year-over-year, and for the full year, that key segment for AI made up more than half of the server market. Nvidia continues to dominate that space with 90% of shipments in the final quarter of 2024.

"The fast pace at which hyperscalers and cloud service providers have been adopting servers with embedded GPUs has fueled the server market growth which has more than doubled in size since 2020 with revenue of $235.7 billion dollars for the full year 2024,” IDC said.

All that growth raises challenges, in particular surrounding the high energy use of data centers amid predictions the AI boom will lead data centre power demands to triple in the US.

Gartner also predicts 40% of data center operators will face power constraints — sending some tech companies to invest in alternative power sources such as nuclear.

"Growing concerns around energy consumption for server infrastructure will become a factor in data centers looking for alternatives to optimize their architectures and minimize energy use," Fernandez predicted.

Winners and losers

Dell and Supermicro remain atop IDC's tracker, but their dominance is increasingly being challenged. Dell held a 7.2% market share in the fourth quarter of 2024, while Supermicro held 6.5%.

Yet despite overall revenue growth, both had lost market share from the previous quarter, from 11.3% and 8.0% respectively.

HPE, IEIT Systems, and Lenovo all shared the next spot on the tracker, each with shares around the 5% mark. The original design manufacturer segment — which can reflect hyperscaler demand — made up almost half of the total revenue (47.3%), while the "rest of market" section nearly a quarter (23.7%), highlighting a significant shift in the market.

Indeed, that marks a big change from the fourth quarter of 2019. In that quarter, the global server market revenue grew just 7.5% year-over-year to $25.4 billion — versus last quarter's $77.3 billion in growth just five years on.

In 2019, ODMs had just a quarter of server revenue and "rest of market" just 16%.

MORE FROM ITPRO

• Why liquid-cooled infrastructure has become mainstream
• Looking to upgrade your servers? Here's our best picks
• What is infrastructure as code (IaC) and what are its benefits?

Microsoft has resolved an issue that saw companies running Windows Server 2022 and Server 2019 unexpectedly upgrading to Windows Server 2025 — despite not paying for the upgrade.

Windows Server 2025 was released earlier this month with new features including upgrades to security in Active Directory, hotpatching for hybrid and on-prem clouds, and more. However, it also came with a few notable bugs, including one that crashed computers.

Shortly after the release, users spotted that older versions of Windows Server were suddenly upgrading to Windows Server 2025.

One employee of a small business in the UK posted on Reddit that Windows 2022 Servers had upgraded themselves to the new version, or were about to do so.

"This obviously came as a shock as we're not at the point to do so for many reasons and the required licensing would not be present," the user wrote.

Unwanted Windows Server 2025 upgrades

Microsoft said in an update that there were two scenarios in which this happened. The first involved devices automatically upgrading to the new OS.

"This was observed in environments that use third-party products to manage the update of clients and servers," Microsoft explained.

The tech giant noted the issue had been mitigated, but also advised users to ensure third-party software designed to manage updates be configured to not automatically deploy them.

Microsoft updated the release under the upgrade classification "optional" — suggesting there was an issue with how the software was classified for admins and third-party software.

That was backed up in a separate post by a technical manager at security and update software company Heimdal, which noted that the "KB" label appeared to be incorrect. Whether that caused the unexpected updates or not, Microsoft has fixed the "KB" issue.

"The Windows Server 2025 feature update was made generally available on November 1, 2024, as KB5044284, which was the same KB number used for Windows 11, version 24H2," Microsoft said in its update.

"This was the KB numbering for both these client and server Windows updates available at that time. Future updates released for Windows Server 2025 and Windows 11, version 24H2 will share the same KB numbers, but will have different release note sites and links."

Regardless of whether third-party management software is to blame or Microsoft itself, plenty of admins had a tough week.

"It's not been a great week at work, too much time lost on this, and the outcome is that in some instances backups have come into play however Windows Server 2025 licensing will have to be purchased for others," the small business employee noted in their post.

The second unexpected upgrade scenario involved a banner displayed on the device's Windows Update page in Settings, offering an upgrade to the new OS.

"This message is intended for organizations that want to execute an in-place upgrade," the company said, adding that the issue had since been resolved.

Recurring issues

Microsoft quickly spotted three bugs in Server 2025 - one that meant English text was shown during installation, even if it wasn't the default language, and a second that limited the OS from running on devices with a high core count, sparking crashes, blue screens and long restarts.

Similarly, a third error popped up saying "boot device inaccessible" in iSCSI environments.

Microsoft has acknowledged all three issues, saying it is working on a resolution for each, and promising a solution in a future Windows monthly update for the core count and iSCSI bugs, and merely more information for the English text issue.

Cloud computing may be all the rage but when faced with ballooning costs, ever more frequent outages and the worrying trend in data breaches, businesses of all sizes are seeing the tangible benefits offered by on-premises servers. With the appropriate supporting infrastructure in place, they can deliver 24/7 data availability and put you in full control of all services and business apps. 

True, your capital expenditure (CapEx) will be higher, but compared to ongoing cloud service subscriptions, your operating expenses (OpEx) over the long term could be significantly lower. You’ll have total equipment ownership and be free to make any customisations you want, whenever you want, store your data in more cost-effective ways, and be in full control of all aspects of security, including data governance.

One of the biggest challenges facing businesses is the sheer range of servers on today’s market, making it important to choose the right one. Over the past twelve months, we’ve lab-tested and reviewed servers ranging from enterprise powerhouses to small business models, and in this guide, we bring you our pick of the best ones.

Best Business Servers 2026

Dell's PowerEdge R770 is one of Dell’s most versatile rack servers yet, as it teams up Xeon 6 power with a high memory capacity and a wide range of storage options. This 2U rack mounter is available in hot and cold aisle configurations, supports air-cooling or direct liquid cooling (DLC) kits, and expansion options include plenty of GPU choices.

Its excellent internal design has room for two Xeon 6 P-Core CPUs up to 86 cores or E-Core models up to 144 cores, and if you choose the former, you can use the 32 DIMM slots to push capacity to an incredible 8TB of fast DDR5 memory. Storage features are outstanding as the R770 can handle up to 24 SAS4/SATA devices, including 8 SFF NVMe SSDs or up to 40 E3.S Gen5 NVMe SSDs.

Dell has wholeheartedly embraced the DC-MHS specification, so you’ll find its fPERC RAID cards and BOSS-N1 controllers are fully compliant. Dell also delivers unbeatable remote management services as its iDRAC10 modular controller presents a wealth of valuable server information, plus essential platform security and links up with Dell’s OpenManage Enterprise and Power Manager, along with its AI-driven AIOps infrastructure management cloud service.

Price: Starts from £6,100 excluding VAT

Read our full Dell PowerEdge R770 review for more information

Enterprises seeking a powerful edge server that can cope with harsh environments will find a lot to like about HPE’s ProLiant DL145 Gen11. Along with support for all twelve of AMD’s EPYC 8004 CPU family, including the six ‘PN’ NEBS-friendly models, this compact 2U system is resistant to extreme shock and vibration and rated for continuous operations in temperatures between -5 and +55 degrees C. 

Its high CPU core potential makes it a good choice for virtualisation at the edge, and it’s ready for AI and ML workloads as it can handle up to three Nvidia L4 Tensor Core GPUs. Maximum memory capacity is 768GB of fast DDR5, and although storage capacity isn’t high, it supports two SATA/NVMe SSDs and has options for up to six EDSFF E3.S SSDs.

Management features are undiminished as it sports an embedded iLO6 chip, which offers strong security, zero-touch provisioning, and full integration with HPE’s GreenLake Compute Ops Management cloud service. The DL145 Gen11 is clearly capable of delivering plenty of processing power to the network edge, and deployment options are extensive as HPE offers desk, rack, and wall-mount kits, so it can happily go where normal servers fear to tread.

Price: Starts from £2,393 excluding VAT

Read our full HPE ProLiant DL145 Gen11 review for more information

Broadberry’s CyberServe EPYC EP1-226T has some big surprises in the storage department as this 2U rack server sports a SupremeRAID SR-1010 RAID solution from GRAID Technology. Hardware RAID cards can present performance bottlenecks when managing high-speed NVMe SSDs, but GRAID Technology neatly overcomes these issues by offloading all RAID operations to a GPU and using out-of-path technology to reduce CPU load.

It certainly delivers on its speed claims as in our lab tests using twelve 3.8TB Micron 7450 PRO U.3 NVMe SSDs, Iometer reported mind-boggling throughputs of 58GB/sec and 21GB/sec for sequential reads and writes, with random operations returning 56.2GB/sec and 20.6GB/sec. Swapping Iometer to a 4K request size delivered equally impressive I/O rates with sequential reads and writes settling at 1.11million IOPS and 591,100 IOPS, while random reads and writes both returned around 1.1million IOPS.

The CyberServe works for us as a powerful general-purpose storage server, as it comes with a 24-core 2.9GHz AMD EPYC 9254 CPU, and the price includes a generous 128GB of DDR5 memory. The server is limited to a maximum of 16 NVMe SSDs, but the SupremeRAID card makes it an ideal choice if you want to get the best from your storage investment.

Price when reviewed: £15,906 excluding VAT

Read our full Broadberry CyberServe EPYC EP1-226T review for more information

First introduced way back in 2010, HPE’s ProLiant MicroServer just keeps marching on with the latest Gen11 model delivering a wealth of new features. Targeting small offices seeking a compact entry-level server, it’s now available with a choice of Intel Pentium Gold, Xeon E-2400 or Xeon 6300 CPUs and supports up to 128GB of DDR5 memory.

It’s slightly larger than the Gen10 Plus model, and HPE has put the extra space to good use by radically improving system board access. Its smart clamshell design allows the upper section of the inner chassis to be swung over through 90 degrees to provide unfettered access – you don’t even need to remove any cables. 

Storage looks good as there’s room for four LFF SATA NHP (non-hot-plug) HDDs or SATA SFF drives using optional converter kits. RAID is handled by an embedded Intel VROC (virtual RAID on CPU) controller, and remote management is a standard feature as the server sports HPE’s excellent iLO6 controller.

Price: Starts from £714 excluding VAT

Read our full HPE ProLiant MicroServer Gen11 review for more information

Most small business servers tend to be black and boring, but Dell’s PowerEdge T160 breaks the mold as it packs an impressive hardware package into a compact and stylish chassis. Presented in eye-catching silver, the T160 has a zinc coating for protection, while its industrial-strength construction and optional bezel filter kit make it a solid choice for near-edge deployments such as warehouses and retail offices.

Available with a choice of Intel Pentium G7400, Xeon E-2400, or the latest Xeon 6300 CPUs, the server supports a healthy 128GB of DDR5 memory and offers an impressive storage potential. The two internal cages support three LFF plus two SFF SAS/SATA devices, and there’s even room below for Dell’s optional BOSS-N1 card, which presents two M.2 NVMe SSDs in hot-plug carriers accessible at the rear. 

Full remote management is included, and although the embedded iDRAC9 controller isn’t the latest generation, its web console presents a wealth of information about system and component status. Small businesses will also appreciate Dell’s embedded LifeCycle Controller as it makes light work of OS installation. 

Price: Starts from £1,172 excluding VAT

Read our full Dell PowerEdge T160 review for more information

Businesses looking for a NAS appliance offering a unified storage solution are faced with a big choice of vendors but OWC (Other World Computing) stands out for several compelling reasons. Comprising 8 and 16-bay rack-mount appliances, its Jupiter Callisto models deliver powerful hardware packages plus a rich set of enterprise-class storage features all offered at competitive prices.

One of the most notable features is the iXsystems TrueNAS Scale operating system at their hearts. Many businesses and enthusiasts will undoubtedly be familiar with TrueNAS Core (formerly FreeNAS) software but the Scale version is aimed at much larger environments such as data centers and delivers a lot more storage features with a sharp focus on scalability.

Both use the highly reliable OpenZFS file system but whereas TrueNAS Core is based on FreeBSD, Scale is built on Debian-Linux. Key features are support for clustering, scale-up and scale-out operations, integral virtualization services, and the ability to run K8s (Kubernetes) or Docker apps.

You don't need to worry about relying only on community support for TrueNAS Scale either. OWC provides full customer support for all aspects of the Jupiter Callisto including deployment and troubleshooting plus the price includes a three-year hardware warranty extendable to five years.

OWC Jupiter Callisto review: Hardware

We review the 8-bay Callisto model which is presented as a 2U rack server and comes ready for action as the price includes eight 4TB Toshiba enterprise-class MG08 SATA HDDs preinstalled in hot-swap carriers. Peek round the back and you'll find a pair of mirrored 480GB OWC Mercury Extreme Pro SATA SSDs dedicated to the OS.

Under the lid, it all looks nice and tidy with the Supermicro X11SPi-TF motherboard sporting a 2.2GHz 10-core Xeon Scalable Silver 4210 CPU mounted by a large active heatsink. TrueNAS Scale memory requirements are more than covered as all Callisto models come with a generous 128GB of DDR4 ECC upgradable to 512GB.

The HDD backplane is hooked up to a Broadcom SAS/SATA PCIe HBA with RAID functions handled by TrueNAS. Our system was supplied with all drives configured as a Z2 array which is the equivalent of RAID6, but you can opt for stripes, mirrors, Z1 (RAID5), or Z3 for triple parity.

A Broadcom PCIe card looks after storage expansion and its dual mini-SAS3 ports can handle up to seven 8 or 16-bay OWC Jupiter Kore disk shelves allowing raw capacity to be boosted to an impressive 2.5PB. The appliance comes with dual embedded 10GbE copper ports and the three spare PCIe slots can be used to add extra 40GbE or 100GbE cards. 

OWC Jupiter Callisto review: 10GbE performance

For NAS performance testing, we created a share on the appliance and mapped it over 10GbE to a Dell PowerEdge T640 Xeon Scalable tower server running Windows Server 2022. We saw excellent NAS speeds with Iometer reporting sequential read and write rates of 9.3Gbit/sec with random rates settling comfortably at 9.3Gbits/sec and 9.2Gbits/sec.

The random test results are of particular interest as these show how efficient the TrueNAS ZFS read and write caching is. You can keep an eye on its memory usage from the web console and during our tests, ZFS caching never consumed more than 2GB.

Real-world performance was equally impressive with copies of a 25GB test file between the NAS and server averaging read and write rates both of 9Gbits/sec. Our 22.4GB backup test folder containing 10,500 small files was secured at a speedy 2.3Gbits/sec while copying our 25GB file to an encrypted share averaged 3.4Gbits/sec.

Moving over to IP SANs, we created a 1TB iSCSI target on a new Zvol and watched Iometer report sequential and random read speeds both of 9.2Gbits/sec. Write speeds weren't quite so good though, as we recorded sequential and random rates of 6.1Gbits/sec and 6Gbits/sec.

OWC Jupiter Callisto review: TrueNAS Scale features

Data protection features are outstanding with ZFS copy-on-write (COW) delivering near unlimited NAS and iSCSI LUN snapshots. These can be easily scheduled from the main web console where you choose a dataset, set snapshot longevity in days, weeks, months, or years, and apply a predefined schedule or create your own.

Data corruption is handled transparently with end-to-end checksums, snapshots can be replicated to remote appliances, and Rsync is also supported. All popular cloud storage providers are on the TrueNAS guest list and after adding our Amazon S3 account credentials, we could create scheduled cloud sync or copy tasks for selected directories.

Space-saving options are plentiful as TrueNAS applies inline compression and you can create datasets with deduplication enabled. NAS share contents can be protected from tampering or deletion by activating WORM (write once read many) during creation and you have five minutes to tweak them before it is enforced.

The appliance can act as a virtualization host and we had no problems creating a Windows Server 2022 VM from an ISO file with it ready for action in twenty minutes. You're spoilt for choice with apps and although they aren't as business-focused as those from Qnap and Synology, you're sure to find many useful ones with 99 currently available on the app discovery console page.

The Callisto was supplied with TrueNAS Scale Bluefin and after testing, we had the opportunity to upgrade it to the latest 23.10 'Cobia' version. This heralds in a wealth of new features including a Linux 6.1 kernel, enhanced hardware support, ZFS block cloning for faster file copying plus the ability to scale to 1,200 drives on a single system for a mind-boggling 25PB storage capacity.

OWC Jupiter Callisto review: Is it worth it?

With the fully-prepped system on review costing £8,991 exc VAT, the Jupiter Callisto represents excellent value. With TrueNAS Scale at the helm, it's easy to deploy and delivers an impressive range of unified storage features with a sharp focus on data integrity.

Overall performance is very good as although IP SAN sequential and random write speeds were a tad below expectations, the Callisto delivered top NAS rates during our 10GbE tests. The Jupiter Callisto is a great choice for businesses that want a powerful and highly expandable storage appliance and OWC seals the deal with its support package and hardware warranty.

OWC Jupiter Callisto specifications (as reviewed)

Businesses usually expect to pay a premium for edge servers but Broadberry's CyberServe Xeon SP1-102 NVMe G4 bucks the trend as it offers a compact solution with an equally small price tag. A chassis depth of only 520mm also makes this 1U rack server well-suited to remote or branch office deployments.

Don't be deceived by its modest dimensions as the CyberServe is capable of delivering a very powerful package. It supports Gen4 Xeon Scalable CPUs, has a mighty memory capacity, and offers a surprisingly good expansion potential.

Showcasing Gigabyte's E163-S30 rack server platform, it's thermally rated to handle CPUs up to a 350W TDP so you can choose just about any model you want. Broadberry opted to keep costs under control by dropping a basic 8-core 1.8GHz Bronze 3408U CPU in our review system but you can go right up to a 60-core Platinum. 

Broadberry CyberServe Xeon SP1-102 NVMe G4: Design

The server's chassis is of solid construction and under the lid you'll find a well-designed interior with cable-related clutter kept to a minimum. At the heart of the CyberServe is Gigabyte's MS33-DC0 motherboard which presents its single CPU socket towards the front with the Bronze module mounted by an impressively large passive heatsink. 

The socket is flanked on each side by a total of 16 DIMM slots allowing the CyberServe to support up to 4TB of fast 4,800MHz DDR5 memory using 256GB 3DS RDIMM modules. The server supports standard RDIMM modules for a maximum of 1.5TB and value looks even better as the price includes a healthy 256GB made up of four 64GB modules.

The server only offers a single Gigabit network port but there's loads of room to expand. Two risers at the rear each provide PCIe Gen5 x16 slots and both have room for full-height, half-length (FHHL) adapter cards. 

There's more as the motherboard has OCP (Open Compute Project) 3 bays on each side of the riser assembly and industry-standard mezzanine cards can be installed from the rear without needing to open the server's lid. As with the standard PCIe slots, these are the faster Gen5 x16 variety and Broadberry offers a big choice of Intel and Broadcom mezzanine cards ranging from quad-Gigabit and 10GbE right up to dual-port 100GbE.

Broadberry CyberServe Xeon SP1-102 NVMe G4: Storage services

Storage features are basic as the server only has room at the front for two SFF hot-plug drive bays. This isn't an issue if you're using the server to provide edge or general networking services and the bays support SATA and NVMe SSD storage devices out of the box.

Our system includes a pair of 2TB Micron 7450 Pro U.3 NVMe SSDs and cabling is kept to a minimum as the motherboard's MCIO (Mini Cool Edge IO) PCIe Gen5 connectors are located right next to the drive cage. Another smart motherboard feature is it comes with Intel's embedded VROC (virtual RAID on CPU) controller and Broadberry includes the standard hardware activation key. 

Obviously, with only two drive bays you're limited to striped or mirrored RAID arrays but the VROC controller supports these for both SATA and NVMe devices. This is one area where Xeon Scalable servers score over AMD as most EPYC-equipped servers don't have embedded SATA/NVMe controllers or integral RAID capabilities.

The motherboard also has a single M.2 NVMe SSD slot that supports 2280 and 22110 cards. This'll come in handy for running a hypervisor and unlike the blue chips, Broadberry allows you to install your own choice of storage devices without invalidating the warranty.

Broadberry CyberServe Xeon SP1-102 NVMe G4: Monitoring and management

If you're planning on accessing the server locally, bear in mind it has a mini-DP port. This shouldn't be a problem if you don't have a suitable monitor as you can pick up an HDMI or VGA adapter from Amazon for less than a tenner.

For remote management, you have the server's embedded AST2600 IPMI controller and dedicated Gigabit port. Gigabyte has redesigned its web interface and replaced the dashboard's gaudy color blocks with a more informative set of tables showing system and component status along with a graph for power consumption.

You can drill down deeper and pull up detailed graphs of all sensor readings and use the controller's PEF (platform alert filter) services to create policies that issue emails or SNMP traps if any exceed their thresholds and carry out actions such as resetting or powering down the server. The server's bank of five dual-rotor cooling fans are a tad noisy but custom fan profiles can be linked to sensors to control their speed and unlike Dell and HPE, you get full OS remote control and virtual media services as standard and not as a chargeable license upgrade.

Multiple Gigabyte servers can be managed and monitored using the free Gigabyte Server Management (GSM) utility. We ran it in the lab on a Windows Server 2019 host and after its network discovery service located the server's IPMI controller, it loaded a basic dashboard showing the status and overall power consumption of all monitored systems.

Selecting one pulls up a hardware inventory and moving to the monitoring section reveals graphs of temperatures, fan speeds, and voltages for up to a year. A separate section shows power consumption graphs and the console provides tools for remote power controls and firmware upgrades.

Broadberry CyberServe Xeon SP1-102 NVMe G4 review: Is it worth it?

There is some blue-chip competition at the network edge but in most cases, their solutions are more specialized. Dell's ruggedized PowerEdge XR11 for example, is aimed at harsher environments such as military use, only supports Gen3 Xeon Scalable CPUs, and is a lot more expensive.

With the system on review costing £3,920, the CyberServe Xeon SP1-102 NVMe G4 delivers an affordable, low-profile rack server solution. It offers plenty of PCIe Gen5 expansion options, supports all the latest Gen4 Xeon Scalable CPUs, and clearly has the power to run a wide choice of applications.

Broadberry CyberServe Xeon SP1-102 NVMe G4 specifications

System

1U Gigabyte E163-S30

Motherboard

Gigabyte MS33-DC0

CPU

8-core 1.8GHz Intel Xeon Scalable Bronze 3408U

Memory

256GB 4,800MT/s DDR5 ECC (max 4TB)

Storage bays

2 x SFF NVMe/SAS/SATA hot-swap drive bays 

Other storage

1 x M.2 NVMe SSD slot

RAID

Intel VROC for SATA/NVMe

Storage included

2 x 2TB Micron 7450 Pro U.3 NVMe SSDs

Network

1 x Gigabit

Expansion

2 x PCIe Gen5 x16 slots, 2 x OCP 3 Gen5 mezzanine slots 

Power

2 x 800W 80 Plus Platinum hot-plug PSUs

Management

Aspeed AST2600 RMC with Gigabit, Gigabyte Server Manager

Warranty

3yrs advanced replacement 

Space-poor small businesses seeking their first purpose-built server will find Lenovo's ThinkSystem ST50 V2 a tempting proposition as it's one of the smallest tower servers we've yet seen. It replaces the four-year old ST50 model and the first thing you'll notice is the new model is even smaller. 

The ST50 chassis has a 25-litre volume and the ST50 V2 drops this to 17 litres with a chassis depth of only 30cms so it'll easily fit underneath a desk. Processing power gets a boost as the ST50 V2 supports all ten members of Intel's Xeon E-2300 CPU family and memory capacity has been doubled to 128GB of TruDDR4. 

Three PCIe expansion slots are still present but the x16 slot has been upgraded to a faster Gen4 version. The smaller chassis does have a negative impact on storage though, as it only supports two LFF and one SFF drive (the ST50 has room for four LFF devices) but the ST50 V2 has an extra slot on the motherboard for an M.2 NVMe SSD.

Lenovo ThinkSystem ST50 V2: Build and internal design

The sturdy chassis is of all-steel construction and the removable side panel can be padlocked shut or protected with a Kensington lock. Its 4U wide chassis can be placed on its side in a rack cabinet but note that the optional rack mount kit includes a tray to place the server on which increases the height by a further 1U.

Upgrades are easy to carry out as the server presents a very tidy interior with clear access to all key components. The front plastic grill can be removed only when the side panel is detached and it has knock-out panels for an optical drive and extra drive bay.

Lenovo offers two types of CPU heatsinks with an active version for chips with a maximum 80W TDP. Our system came with a 6-core 3.5GHz Xeon E-2386G and its higher 95W TDP is handled by a passive heatsink with heat pipes routed to a finned radiator mounted in front of the rear cooling fan.

The rear fan is partnered by an 8 cm diameter fan at the front and noise levels are incredibly low. We could barely hear the server when it was running making it a great choice for small offices that value tranquillity.

Lenovo ThinkSystem ST50 V2 review: Storage conundrums

Base systems come with a standard LFF drive bay in the chassis base and we recommend specifying the second top-mounted LFF bay. It's never a good idea running a server OS on a single drive and this allows you to create a mirrored RAID array to protect against a drive failure.

The server's integral Intel VROC (Virtual RAID on CPU) controller supports SATA drives and software-managed RAID0, 1, 10, and 5 arrays. The latter array type is possible but only if you request a 'Special Bid' order comprising three SATA SFF SSDs.

The M.2 NVMe SSD is mounted at the base of the motherboard but you can ignore the M.2 slot next door as this is for a Wi-Fi module which is not currently available. There's just enough room below the M.2 card for a standard SFF SSD mounting tray.

The server and VROC controller only support 'simple-swap', or cold-swap, SATA drive bays. Hot-swap capabilities are not an option but Lenovo does offer its entry-level RAID 5350-8i RAID adapter if you want higher performing SAS drives.

Lenovo ThinkSystem ST50 V2: Management and monitoring

If you're planning on only accessing the server locally, bear in mind the ST50 V2 provides dual DisplayPort connectors and doesn't have a VGA port. We got around this by using Lenovo's DisplayPort to VGA monitor cable (Part: 57Y4393) which Amazon sells for £32.

As with its predecessor, the ST50 V2 isn't blessed with Lenovo's XClarity Controller (XCC) management chip. Instead, you get Intel's built-in AMT (active management technology) 15.0 for managing and monitoring the server.

This is activated by pressing Ctrl-P during the boot-up sequence and security is tight as the default password must be changed after first use. You can decide which features to enable but your choice of CPU will determine which ones are available.

Choose a dual-core Pentium G-series chip and you'll only get Intel Standard Manageability (ISM) which is a subset of AMT and offers hardware inventory, power controls, health monitoring, event logging, and alerting. AMT is supported by all Xeon E-2300 CPUs but to enable AMT's KVM remote control, you'll need to specify one of the six 'G' models which have embedded Intel UHD graphics.

The AMT chip can be accessed from a web browser and to use KVM we installed Intel's free Manageability Commander utility on a Windows 10 host. This provides plenty of information on hardware and events with an option to run remote desktop sessions to the server.

For assisted OS installations, you download Lenovo's XClarity Provisioning Manager Lite ISO file and create a bootable USB stick – Rufus works fine. This loads an interface for creating RAID arrays or updating firmware and we used its guided OS install option to deploy Windows Server 2022. Unlike the full-fat XCC, it doesn't support virtual media so your OS image must be presented on a USB stick or DVD.

Lenovo ThinkSystem ST50 V2: Is it worth it?

With support for a Xeon E-2300 CPU and up to 128GB of DDR4 memory, the ThinkSystem ST50 V2 has plenty of horsepower to run a wide range of business apps. It lacks Lenovo's full-strength XCC management controller although Intel's AMT and Lenovo's XClarity Provisioning Manager Lite should be sufficient for most small businesses.

Storage features are more restricted than the original ST50 but a plus point is the tiny chassis will fit almost anywhere. Its silent runnings will satisfy your inner librarian and with prices starting as low as £350, it's an affordable option for small businesses and remote offices.

Lenovo ThinkSystem ST50 V2 specifications

Don't be deceived by this rack server's modest dimensions as Lenovo's ThinkSystem SR250 V2 offers a surprisingly good hardware package in its short-depth chassis. Aimed at growing SMBs that want a flexible on-premises solution, this compact 1U single-socket server also looks good value with prices for entry systems starting at around £1,100.

On paper, it looks a lot more capable than competing single-socket SMB rack servers from other vendors. Taking over from the four-year-old SR250 model, it supports all ten members of Intel's latest Xeon E-2300 CPU family, doubles maximum memory capacity to 128GB, and puts plenty of Gen4 PCIe expansion slots on the table.

Storage features don't see any significant enhancements but none are needed as the original SR250 and the new V2 model both support four LFF or up to ten SFF drives. The only other entry-level Xeon E-2300 rack server we can think of that's capable of matching this high storage density is Fujitsu's Primergy RX1330 M5.

Lenovo ThinkSystem SR250 V2 review: Storage choices

Lenovo offers three chassis variants with entry systems sporting four LFF hot-swap drive bays. For higher device densities, you can choose an eight SSF bay version or pull all the stops out and go for the ten SSF bay model.

A minor change to storage from the older SR250 is the removal of the AnyBay backplane which allowed two NVMe SSDs to be used in the last two bays. You can still use one NVMe SSD in the SR250 V2 but only with the four SATA LFF 'simple-swap' (cold-swap) configuration.

RAID options are extensive as all models include an embedded Intel VROC (virtual RAID on CPU) controller which supports eight simple-swap SATA drives and software-managed RAID0, 1, 10, and 5 arrays. Hardware RAID controller choices include the ThinkSystem RAID 5350-8i SATA/SAS3 cache-less PCIe adapter provided in our review system with the server also supporting the 9350 models which add cache protection and RAID6 arrays.

You can save all your front bays for data storage as the SR250 V2 supports Lenovo's optional dual-slot M.2 SATA SSD adapter card. It links up with the VROC chip to provide a mirrored array for running an OS or hypervisor but, unlike Dell's BOSS cards, it will take up a PCIe slot.

Lenovo ThinkSystem SR250 V2 review: Hardware

Cracking the lid reveals a tidy, well-designed interior with the supplied 6-core 2.9GHz Intel Xeon E-2336 CPU located on the right-hand side and topped off with a large passive heatsink. Snuggling up alongside are four DIMM slots and Lenovo opted to show off our system's capabilities by populating them all with 32GB TruDDR4 modules.

The server comes with dual embedded Gigabit network ports and there's room to add more as the two available PCIe Gen4 slots support Lenovo's dual and quad-port Gigabit and 10GbE cards plus its dual-port 10/25GbE SFP28 adapters. The third PCIe slot is reserved for Lenovo's ThinkSystem RAID cards and note that if you try to save money by specifying a cheaper dual-core Pentium G-series CPU, the PCIe slots will drop to Gen3 speeds.

Four cold-swap fans behind the drive backplane handle all chassis cooling and general noise levels are low enough for it suitable for a small office. If you add Lenovo's optional Nvidia Quadro T1000 GPU card, bear in mind it will cause the fan speeds and noise levels to ramp up – not hugely, but noticeably.

Power supply choices include a single fixed 300W PSU for lower-specified systems. The other option is a dual bay for 450W hot-plug PSUs and you'll need these if you specify an 8-core Xeon E-2300 CPU, the 10-bay SFF chassis, or the GPU card.

Lenovo ThinkSystem SR250 V2 review: Server management

A clever feature of the ten-bay model is Lenovo has sufficiently reduced the width of its drive carriers to free up enough room to the right for its operator panel which offers a bunch of status LEDs plus USB 3.2, USB 2, and VGA ports. The USB 2 port plays an extra management role as connecting a mobile to it and pressing the server's ID button for more than three seconds allows you to use the XClarity Mobile app to directly view the system's status.

For general remote management, the SR250 V2 sports the same XClarity Controller (XCC) chip and dedicated Gigabit port you'll find on all other ThinkSystem servers. It presents a smart web console offering plenty of details about system and hardware status, utilization graphs for CPU, memory, and power, hardware inventory, and access for managing RAID arrays. Lenovo included an XCC Enterprise license with our server which enables remote KVM, virtual media services, power capping, and remote OS deployment tools.

A feature we found very useful is the embedded provisioning manager and its Effortless Install option. After mapping a Windows Server 2022 ISO file to the server, we nipped out for a coffee and came back 30 minutes later to find the OS ready for action with all required drivers installed.

We run Lenovo's XClarity Administrator in the lab as a Hyper-V VM and after discovering the server's BMC network port, it allowed us to remotely monitor it. The free version includes a 90-day trial of all features but after this has expired you'll need to cough up for a Pro licence for each managed device which costs around £16 each.

Lenovo ThinkSystem SR250 V2 review: Is it worth it?

Lenovo's ThinkSystem SR250 V2 is a very affordable choice for space-poor SMBs and remote offices seeking a powerful and compact 1U rack server. A big choice of Xeon E-2300 CPUs and a high memory capacity makes it very versatile, the embedded XCC controller provides some great server management features and it stands out from the crowd of entry-level rack servers with its ability to present up to ten SFF drive bays.

Lenovo ThinkSystem SR250 V2 specifications 

Aimed at businesses seeking a rack-optimized powerhouse, HPE's ProLiant DL360 Gen11 targets a diverse range of workloads including databases, virtual desktop infrastructure (VDI), edge acceleration, exploratory data analysis (EDA), computer-aided design (CAD) and intelligent video analytics. Its slim-line 1U chassis sees a significant redesign over the five-year-old DL360 Gen10 and a sharp focus on thermal capacity allows it to support the latest Gen4 Xeon Scalable CPUs right up to Intel's top-dog 60-core Platinum 8490H.

There's plenty more on the table as the server squeezes in 32 DIMM slots allowing it to support a maximum 8TB of HPE's DDR5 SmartMemory. Expansion potential looks good as along with support for up to three PCIe Gen 5 expansion slots, it also has room for two OCP 3 mezzanine cards.

Storage sees a shake-up as you can order chassis with a choice of LFF, SFF, NVMe U.3 SSDs and even have up to twenty E3.S EDSFF Gen5 NVMe SSDs in dual CPU configurations. A recently added bracket enablement kit brings GPU support into play too, with options to fit Nvidia's A2 accelerator cards.

HPE ProLiant DL360 Gen11 review: Build and design

The server looks good value with prices starting at around £2,200. Designed to get you out of the starting blocks, the entry-level P60735-421 model sports a 12-core 2GHz Xeon Scalable Silver 4410Y CPU, 32GB of DDR5 SmartMemory, four LFF drive bays and a single 800W hot-plug PSU.

HPE specced our review system with a CTO (configured to order) chassis furnished with the same Silver CPU and 64GB of DDR5 memory. From our experiences, the Silver CPUs are the sweet spot for general-purpose servers as the three Gen4 models offer from ten to twenty cores and base speeds up to 2.7GHz while their maximum 165W TDP doesn't incur any chassis thermal restrictions.

The chassis is solidly built and removing the lid reveals a very tidy interior with the two CPU sockets positioned towards the front of the motherboard and flanked on each side by banks of eight DIMM slots. Cable-related clutter is kept to an absolute minimum and all critical components are easily accessible for maintenance and upgrade maneuvers.

Cooling is handled by a bank of hot-plug fans behind the drive backplane and another advantage of the Silver CPUs is they only require the standard kit of five fan modules so noise levels are pleasingly low. Specify CPUs with TDPs between 186W and 270W and you must move up to the seven-fan kit while HPE offers closed-loop and direct liquid cooling options for higher TDP CPUs.

HPE ProLiant DL360 Gen11 review: Management services

The DL360 Gen11 sports the latest iLO6 embedded controller which offers a wealth of remote management features and has a sharp focus on platform and supply chain security. Along with Secure Start and 'silicon root of trust' firmware fingerprinting, the iLO 6 SPDM (Security Protocol and Data Module) authenticates hardware components, flags up any 'hostile' devices, and uses policies to stop the boot process.

For general management, the iLO6 web interface is packed with status information on all the server's critical components. All ProLiant servers include a Standard license with an optional Advanced version enabling extra features such as real-time power and temperature metering, full OS remote control, and virtual media services.

We run HPE's OneView application on the lab's Hyper-V host which provides a central web console for all our HPE systems and options to monitor hypervisors, storage arrays, pools, and switches. We also run HPE's iLO Amplifier Pack on the same Hyper-V host which provides a separate web console for discovery, inventory, and compliance reporting for up to 10,000 Gen8, 9, 10, and 11 servers. 

Our iLO Amplifier Pack app is also linked up with our HPE InfoSight cloud account so we can use its web portal to view details such as critical server health issues, general wellness, power status, and installed hardware. Last, but certainly not least, is HPE's GreenLake Compute Ops Management cloud service which presents a portal for viewing your entire infrastructure, managing server deployments, and automating system lifecycle compliance policies.

HPE ProLiant DL360 Gen11 review: Storage features

The DL360 Gen11 is available in four storage configurations with base models coming with four hot-swap front LFF drive bays. We have the eight SFF model on review which leaves room for HPE's optional media bay and dispensing with this frees up room for two more SFF bays.

The other alternative is HPE's 20-bay EDSFF option which presents the devices in five banks of four across the front. Naturally, this requires a special backplane fitted but cabling is easy enough as the motherboard's eight embedded slim-SAS ports are neatly arranged across the front and as close as possible to the backplane.

HPE's TriMode backplanes bring SAS and NVMe device support into play and RAID options are plentiful with the standard embedded Intel VROC (Virtual RAID on CPU) supporting SATA drives plus stripes, mirrors, and RAID10/5 arrays in software. A wide choice of RAID cards is also available and our system was supplied with an MR216i-o Gen11 card nestling in the second OCP 3 slot.

HPE's NS204i-u Gen11 boot-optimized storage device is worth considering if you want to keep all your front drive bays free for data storage. Similar to Dell's BOSS-N1, it comes with two hot-plug 480GB M.2 NVMe SSDs, doesn't require a PCIe slot (although it does reduce the maximum slot count to two), and provides mirrored redundant storage for running an OS or hypervisor.

HPE ProLiant DL360 Gen11 review: Is it worth it?

HPE's ProLiant DL360 Gen11 is perfectly poised for rack-dense applications that need plenty of processing power. This 1U rack server is well-designed, it teams up great storage features with a high expansion potential and the smart iLO6 controller unleashes a wealth of management and monitoring services.

HPE ProLiant DL360 Gen11 specifications 

The searing heat of the Nevada desert sun during Pure Accelerate 2023 came with a pinch of irony. It offered a fleeting glimpse into what it might be like being a storage drive in a data center rack unit; roasting hot and gasping for the cool embrace of a power-hungry air conditioning unit. 

Heat has long been an unavoidable reality in the data center. A myriad of spinning disks guzzle power and generate heat for every second they’re active – and they need to remain active. It’s the operator’s job to ensure the temperature is moderated efficiently to avoid dreaded failures. 

But “disks are done”, as Pure Storage CEO Charlie Giancarlo said, and the company predicts by 2028 no new hard disks will be sold, with enterprises shifting to more modern, efficient drives. 

This shift forms part of a natural evolution in the industry, attendees heard, with Giancarlo comparing it to the transition from vinyl records to CDs or VHS and DVDs to streaming services in the last two decades. 

“It only took five or six years for DVDs to completely take over the market from VHS,” he said. “And if we look at 2006, within five years, Netflix, which was sending out DVDs – versus Blockbuster that started out on VHS – completely changed in terms of share. Five years later in around 2011, Blockbuster went bankrupt.”

Making the case for sustainable data storage

The sustainability benefits of flash storage versus legacy disk-based storage make the format highly compelling. It was a recurring theme throughout Pure Accelerate 2023, with several sessions highlighting the growing need for the storage industry to adopt more environmentally-friendly practices. Globally, data centers account for between 1-2% of all global electricity consumption, exacting a significant hit on the environment and power grids.

Traditionally, a big reason for this has been the use of those pesky inefficient disk drives spinning endlessly. The heat generated by these drives demands intensive cooling, which in itself results in added costs and energy consumption. 

In fact, anywhere between 30 to 55% of data center energy usage is dedicated to powering cooling systems, research shows, equating to a significant chunk of overall power consumption. 

Pure is highly vocal on the fact that its own products can unlock up to an 80% reduction in power consumption. Flash drives also provide up to a ten-fold boost to power and space efficiency compared to disk drives. 

That’s not exclusive to Pure products, either. NetApp, a competitor in the space and another industry stakeholder bullish on the potential of flash, reports similar performance metrics.

With that in mind, it seems evident that flash should be the natural go-to for organizations conscious of their environmental impact. But if flash can deliver marked energy efficiency improvements, then why isn’t there widespread adoption as of yet?

IDC research shows more than 90% of exabytes in cloud data centers are still stored on HDDs, with the remaining 10% stored on SSDs or flash. 

Cost, Pure conceded, has been a traditional hurdle for organizations. As various speakers noted, however, enterprise data storage and hyperscaler data centers are also among the “last bastions'' of hard disk usage – and both segments account for a significant portion of global data storage. 

If that is the case, then it’s clear there is still a long way to go until flash storage adoption is ubiquitous and global businesses can begin mounting a serious offensive to reduce energy consumption. 

Data center e-waste is a growing concern

Components used to create the disk drives present at many sites also contribute to the seemingly endless amount of e-waste that scars the earth. 

Shawn Rosemarin, VP for customer engineering at Pure Storage, noted the e-waste aspect of current hard disk use in the data center industry should be a huge cause for concern.

Research from the Circular Drive Initiative (CDI) shows around 90% of hard disks are destroyed when they reach end-of-life, with the majority of drives having just a five-year lifespan. This endemic practice of destroying hard disks adds to the global stream of nearly 54 million tons of e-waste annually.

Flash, however, boasts greater longevity than hard disk and, from a mechanical standpoint, is a far more simplistic design than its predecessors. 

Rosemarin noted that flash storage produces 85% less e-waste compared to disk, which if taken at face value could represent a step change in tackling this global issue. But again, this is dependent on how rapidly adoption continues to grow. 

Data centers and energy crises

Flash storage doesn’t represent a total silver bullet for energy efficiency and carbon reduction for the data center industry, however.

Despite notable benefits highlighted at Pure Accelerate, it is clear that there are significant future challenges ahead - especially when one considers the current state of the global energy market. 

Research from the European Commission projects that data centers will account for 3.2% of all electricity by 2030, marking an increase on current levels and a significant surge since 2018. 

The power demands of data centers have reached such a point that, according to a report from the Financial Times last year, London housing developments were put on hold because there simply wasn’t enough electricity to meet commercial and residential demands. 

These straining demands placed on electricity grids were among one the key challenges Rosemarin highlighted at the conference - and it’s clear this is an issue he expects to continue. 

“We have for the first time in our existence an electricity crisis with no end in sight,” he said. “We’ve already got countries telling public cloud providers ‘we don’t want your tax dollars, we have no more electricity’.”

“And so if we’re only going to turn more GPUs and TPUs and CPUs on, where are we going to find the power to do this? 

IDC research also shows that data volumes are expected to continue growing over the next three years - to around 221ZB of data generated globally. 

With this in mind, it’s clear that flash storage adoption will require a serious acceleration in adoption rates to deliver the power-related benefits it’s being hailed for. 

The performance and capacity advantages are clear, but without widespread adoption, the environmental impact of data centers running largely on hard disks will remain a lingering problem.

An optimistic outlook for flash adoption

Despite obvious challenges, Pure Storage appeared highly confident flash represents “the future” of data storage. 

CMO Matt Burr described the days of hard disks as a “dark age” and believes the “future is flash”. It's a bold claim, but one supported by current statistics. 

Research shows the all-flash array market will grow at a CAGR of around 26.3% over the next five years, representing a serious shift toward broader adoption of flash storage. 

This mood of optimism was prevailing through Pure Accelerate 2023, but while the data storage industry may be leaving the "dark ages" and entering its enlightenment period, key hurdles still lie ahead. 

With AMD's latest EPYC 9004 'Genoa' CPUs at the helm, HPE's ProLiant DL385 Gen11 is ready to deliver some serious acceleration to enterprise workloads. This rack server supports all the EPYC 9004 2P models including the 2.4GHz 96-core EPYC 9654 and with two CPU sockets up for grabs, it can present up to 192 physical cores in 2U of rack space.

Naturally, 4,800MT/s DDR5 memory is on the table and the DL385 Gen11 supports up to 6TB in dual-processor configurations. Expansion potential is equally impressive as the server can have up to eight PCIe Gen5 slots and the motherboard has two OCP 3.0 edge connectors as well.

Businesses seeking a high storage capacity will also find a lot to like, as the DL385 Gen11 offers a remarkable range of storage options including U.3 NVMe and Gen5 EDSFF devices. The chassis is also accelerator optimized and supports up to eight single-width or four double-width GPU cards.

HPE ProLiant DL385 Gen11 review: Design and expansion

HPE supplied us with its Base model for review which costs less than £6,000 and gets you out of the starting blocks with a solid hardware package. This includes a 16-core 3GHz AMD EPYC 9124 CPU, 32GB of DDR5 SmartMemory, an 8-bay SFF front drive cage (no storage included) managed by an MR408i-o Gen11 OCP 3 RAID card, a quad-Gigabit OCP 3 network card and one 800W FlexSlot PSU.

Internal design is impressively tidy as the two CPU sockets are located towards the front of the motherboard with our EPYC 9124 topped off with a very large passive heatsink. You can choose any of the 14 EPYC 2P CPUs and a total of 24 DIMM slots allows it to be upgraded to 6TB although 256GB modules currently command a very high price.

Cooling is handled by a bank of six hot-plug fans behind the drive backplane with our system sporting a set of standard modules. Specify CPUs with TDPs between 240W and 300W and you'll need high-performance heatsinks and fans while CPUs above 300W require the max-performance versions.

Our system includes a primary riser cage with a single PCIe x16 Gen5 slot but this can be replaced with a three slot version and HPE offers secondary and tertiary riser cages with a choice of slot counts and a key feature is they are all the PCIe Gen5 variety. The OCP 3 slots are located under the primary and secondary riser cages and if you want more than Gigabit, HPE offers dual 10GbE, dual and quad 10/25GbE, and even dual 100/200GbE OCP 3 cards.

HPE ProLiant DL385 Gen11 review: Storage bonanza

You're spoilt for storage choices as the DL385 Gen11 supports an incredible range of interfaces, devices, and form factors. Drive count potential is one of the highest we've seen in a 2U rack server as you can specify up to 20 LFF HDDs or 34 SFF HDDs/SSDs.

Using HPE's storage boxes, base systems start with a 4-LFF or 8-SFF front drive box and to increase capacity, you add more boxes along with the appropriate backplanes and RAID controllers. Starting with an 8-bay SFF front box, our system accepts two more at the front and, at the cost of an expansion riser, another 2-bay box at the rear.

The drive count can be boosted further with a 4-LFF or 8-SFF midplane box which sits over the CPUs, replaces the air shroud, and requires the high-performance fan kit. The SFF midplane box supports SAS, SATA and U.3 NVMe drives and both types require lower CPU heatsinks which introduce a 300W TDP restriction although this still allows 84-core EPYC 9634 CPUs to be specified.

Whichever model you choose, you'll need a RAID card as all Gen11 models have dispensed with the onboard SR100i Smart Array controller and another new feature is HPE's NS204i-u Gen11 card. Similar to Dell's BOSS-N1, this fits in a dedicated slot and presents two hot-plug M.2 NVMe SSDs at the rear which can be mirrored and used as a boot device.

HPE ProLiant DL385 Gen11 review: Management

The new iLO 6 embedded controller delivers a wealth of remote management and monitoring services and introduces HPE's trusted security features. Building on iLO 5 features such as Secure Start and 'silicon root of trust' firmware fingerprinting, the iLO 6 SPDM (Security Protocol and Data Module) authenticates all hardware components and optional PCIe expansion cards.

If SPDM spots an unauthenticated component or the addition of a 'hostile' device, it flags this up as a risk in its security log and can use a policy to stop the boot process. RedFish certificate and signature management works with HPE's Secure Boot and the iLO 6 RESTful API to ensure components run during startup are trusted while the PLDM (Platform Level Data Model) service ensures you can safely update device firmware from any remote location without the need to shut down or reboot the server.

Management applications and services are in abundance with HPE's OneView providing a central console for all your HPE systems and can also monitor hypervisors, storage arrays, pools, and switches. We run HPE's free iLO Amplifier Pack on our Hyper-V server which provides a separate web management console for discovery, inventory, and compliance reporting for up to 10,000 Gen8, 9, 10 and 11 servers.

After linking up the iLO Amplifier Pack app with our HPE InfoSight cloud account, we could view details such as critical server health issues, power status, and installed hardware from its web portal. And then there's HPE's GreenLake Compute Ops Management cloud service which presents a portal for viewing your entire infrastructure, automating system lifecycle compliance policies, and managing server deployments.

HPE ProLiant DL385 Gen11 review: Verdict

With AMD's EPYC 9004 CPUs in the driving seat, HPE's ProLiant DL385 Gen11 delivers an unprecedented core density. Add support for up to 6TB of DDR5 memory, an impressive storage capacity, PCIe Gen5 expansion options, multi-GPU capabilities, and a wealth of management and platform security services and you have a versatile rack server that's clearly capable of running virtually any enterprise workload.

HPE ProLiant DL385 Gen11 specifications

The amount of data being generated at the network edge is increasing so rapidly that it makes more sense for enterprises to move servers out there as well so they can process it locally instead of funnelling it all back to the data centre. Lenovo moved into edge computing a couple of years ago and its latest ThinkEdge SE450 Edge Server aims to put some serious processing power in the smallest of spaces.

It's powerful enough to satisfy a wide range of edge applications as this 2U rack server supports a Xeon Scalable Gen 3 CPU. It has the thermal capacity to handle 205W TDP CPUs so it can be specified with a core-heavy Platinum model if required.

The SE450 is designed for harsh environments as the chassis is shock and vibration resistant with the optional lockable security bezel incorporating a dust filter. Standard models are cleared for continuous operations in ambient temperatures of between 5 and 45 degrees C while the NEBS Level-3 and ETSI compliant models can handle extremes of -5 and 55 degrees C for up to 96 hours.

Even better, the standard chassis measures a mere 300mm deep allowing it to go into spaces where standard rack servers can't. This model supports four half-length (HL), full-width (FW) expansion cards with another version accepting four FL cards and increasing depth slightly to 360mm.

Lenovo ThinkEdge SE450 Edge Server review: Design and features

In common with most edge servers, the SE450 employs a reverse system design with all expansion slots, PSU sockets and network ports accessible from the front. System cooling is handled by six dual-rotor, hot-plug fans at the rear which we found to be a little on the noisy side with sound levels at one metre in front measured at around 55dB.

For such a small chassis, Lenovo has done a fine design job with the CPU socket situated in the centre, topped off by a passive heatsink and flanked on each side by five DIMM slots. Our system came with a generous 512GB of TruDDR4 memory which can be boosted to a healthy 1TB and the server also supports Intel Optane Persistent Memory (PMEM) 200 modules.

On each side are expansion riser cards both with two PCI-E Gen 4 16x slots. Along with a good range of Gigabit, 10GbE, 25GbE and 100GbE network cards, you can fit up to four single-width Nvidia A2 GPU cards in the 300m chassis and two double-width Nvidia A30 cards in the 360mm deep version.

A vertical OCP 3 module sits to the far left and our server was supplied with an Intel X710-T2L dual 10GbE adapter. Next to this, you can fit an optional WLAN module and bring secure Wi-Fi 6 connectivity into play.

Lenovo ThinkEdge SE450 Edge Server review: Storage options

The SE450 offers a good range of storage features with an internal tray at the front supporting up to four 7mm or two 15mm devices and these can be SATA or NVMe models. Some patience is required when fitting drives in the tray due to the limited amount of access space.

If you're prepared to lose the right-hand expansion riser, you can replace it with a dual-drive SFF hot-swap bay which also handles SATA and NVMe drives. This is only available with preconfigured and CTO (configure to order) models and not as a field upgrade.

In fact, the server supports a total of six SATA and four NVMe devices so you can still use the drive tray, have a mix of storage types and use the on-board Intel VROC for software-managed RAID arrays. SAS3 support is also a possibility with the SE450 supporting Lenovo's ThinkSystem RAID 5350-8i RAID card and its basic 4350-81 HBA.

The server also has room for a dual slot M.2 SSD module which is tucked away next to the PSU bay, supports SATA or NVMe SSDs and can use the Intel VROC chip for creating a mirrored RAID array. Choose the ThinkEdge SE450 with Security Pack model and any physical interference such as opening the lid, can be linked with actions including automatic power down or full disk encryption.

Lenovo ThinkEdge SE450 Edge Server review: Remote management

The SE450 is endowed with the same full-strength XClarity Controller (XCC) and dedicated Gigabit port as all other ThinkSystem servers which can be accessed at boot up where it presents a provisioning manager. The XCC Standard version provides features such as remote power controls, sensor monitoring and inventory while the Advanced upgrade adds remote KVM, virtual media services, remote OS deployment and real-time performance graphs.

The provisioning manager can be used to configure the server, create RAID arrays and deploy an OS. We used the XCC remote console to mount a virtual Windows Server 2022 ISO and chose the 'Effortless Install' option which lived up to its name by loading the OS in 45 minutes with no further intervention required from us.

Centralised remote management is provided by Lenovo's XClarity Administrator which we run in the lab as a VMware VM. After adding the SE450 to its console, we could remotely monitor it, employ the image library for firmware updates plus OS deployments and access the Administrator host from the XClarity Mobile iOS app to remotely access the Administrator host and monitor our server from an iPad.

Lenovo ThinkEdge SE450 Edge Server review: Verdict

The ThinkEdge SE450 Edge Server looks good value as a starting price of around £2,920 gets you the 300mm chassis filled with a 10-core 2.3GHz Xeon Scalable Silver 4310T, 16GB of TruDDR4 and two 240GB SATA SSDs. Our review system does cost about £9,141 but this delivers a hugely more powerful package clearly capable of handling demanding edge related workloads.

The SE450 impresses with design too, as it packs a remarkable amount of hardware into a chassis that's barely a foot deep. This makes it highly versatile as it can deliver a powerful processing package in places where normal servers simply aren't practical.

Lenovo ThinkEdge SE450 Edge Server specifications

Amazon Web Services (AWS) has announced Lambda SnapStart, an update to the Lambda serverless compute service that greatly reduces application start latency.

SnapStart aims to speed up the cold start times of applications through the use of snapshots which can run multiple execution environments.

Once enabled within a Lambda function, Lambda SnapStart will begin its optimisation, storing an encrypted snapshot of the memory and disk state in the cache to be called upon at a later time. Snapshots that have not been called for 14 days are removed from the cache memory.

AWS Lambda runs functions within a secure execution environment that cycle through an initialisation (Init), invocation (Invoke), and shutdown phase. Running through these every time code is called upon can be slow and lead to increased costs.

With Lambda SnapStart, each snapshot can be retrieved and invoked without the need for the Init phase, and populates the execution environment in chunks as required. This results in far faster load and execution times, and AWS claims Lambda SnapStart can enable Java functions to start up ten times faster.

“Initialising the runtime for some languages can be expensive,” said Jeff Barr, chief evangelist at AWS, in a blog post.

“For example, the Init phase for a Lambda function that uses one of the Java runtimes in conjunction with a framework such as Spring Boot, Quarkus, or Micronaut can sometimes take as long as ten seconds - this includes dependency injection, compilation of the code for the function, and classpath component scanning.”

In the same post, Barr gave an example of the service in action, in which he reduced the initialisation of an example function from just over 6 seconds to under 200 milliseconds. This is possible through the restoration of the snapshot saved in the cache, and the subsequent invocation of the function handler.

For all its benefits, the introduction of Lambda SnapStart does necessitate any changes to existing developer practice. Code that has relied on generating content during the Init phase might require alteration, like pseudo random number generators, for example, as the restoration phase does not provide the unique conditions of initialisation.

With this in mind, AWS has updated functions such as OpenSSL’s RAND_BYTES function so that it works in a SnapStart environment, as well as similar functions within Java and Amazon’s Linux.

Similarly, AWS has warned developers to check over code that downloads reference data during the Init phase, as this can be made stale through SnapStart’s caching.

AWS Lambda is the firm’s serverless computing service, which allows companies to run code without the need to provision or manage servers in-house. Developers can use it to run apps and focus entirely on their maintenance, without the additional burden of servers or hardware.

The announcement was made at the firm’s annual conference AWS re:Invent, held in Las Vegas. SnapStart is available now for Lambda customers at no extra cost, in the Asia Pacific, Europe, US East, and US West regions.

Twitter systems are safe from collapse in the immediate future due to years of infrastructure planning, according to a senior engineer who left the platform in August.

Matthew Tejo, a former Site Reliability Engineer (SRE) at Twitter, explained in a blog post that much of his career at the firm was spent automating systems where possible, and disaster planning where it was not, and that the platform can continue to function providing there are no major changes to the systems in place.

The explanation of how Twitter's infrastructure was designed comes after members of the tech industry questioned whether Twitter would be able to run after new CEO Elon Musk fired large portions of engineering staff.

Tejo said that Twitter relies heavily on cache memory to handle traffic, keep response speeds low across the website, and massively reduce overall server costs.

These caches are then run on the Aurora framework, itself encompassed on the open source Apache Mesos project. While Aurora allocates applications to servers, Mesos aggregates servers, removing them in the case of breaks.

As Mesos is not capable of detecting all hardware issues, Twitter relies on manual monitoring from its IT department to check for problems such as bad disks. If one is found, repair workers in the data centre are automatically sent to rectify the problem.

The small number of Twitter’s remaining workforce - believed to be just 20% of its peak following the most recent round of resignations - could prove problematic, as the same amount of work now has to be completed by fewer engineers.

However, Tejo also revealed that at any given time, Twitter has two concurrently-running data centres capable of handling a total failure of the site, with each capable of running all the core services on the platform. This means that Twitter constantly has 200% capacity, for use in worst-case scenarios, and therefore is incredibly unlikely to die through a lack of server resources.

Twitter also uses custom tools to ensure that servers are safely distributed from the moment they are allocated: “Those tools make sure the team doesn’t have too many physical servers on a rack and that everything is distributed in a way that won’t cause problems if there are failures,” said Tejo.

Unknown problems with the infrastructure, or changes to it made in the wave of alterations brought in by new Twitter CEO Elon Musk, could still destabilise the platform. Reflecting on the amount of effort that has gone into making Twitter at least partially self-sustaining, Tejo nevertheless acknowledged that he is “sure there are some bugs lurking somewhere”.

In the immediate aftermath of Musk taking over, Reuters reported that Musk was seeking to make $1 billion in infrastructure cuts in the coming months. The source reporters spoke to indicated that $1.5 to $3 million in server and cloud services costs had been identified as unnecessary, suggesting that the excessive safety redundancies which Tejo helped establish might not be maintained.

“I don’t want to be using systems or services that are hurriedly assembled under extreme duress, standards will slip, data will get lost,” Jeff Watkins, CPTO at xDesign told IT Pro.

“Worse still is that the likely outcome will be a great brain drain. As a result, the remaining team will likely not be the A-team.

“So the user data impact could be bad, but Twitter isn’t just used by the tweeting users via the website and mobile applications, it also has some interesting side-effect usage through its APIs, like detecting downtime in systems (from tweets mentioning the particular company). Destabilising what has become almost a social shadow-IT could have some unexpected consequences on a global scale.”

It remains unclear if Twitter's infrastructure will sustain the platform in the long term with fewer engineers working on it. Despite the excess server resources available, bugs are rampant in software and experienced engineers are required to address them to ensure the smooth running of services.

Twitter has undergone a period of rapid changes since Elon Musk completed his acquisition of the platform on 27 October. In the weeks since, a number of senior figures at the company left their roles, half its employees were fired overnight amidst chaotic scenes of workers being locked out of their emails, and a large number of remaining workers responded to Musk’s demands of harsher work conditions with a resignation ‘revolt’.

On Monday, The Verge reported that Musk made huge cuts to Twitter staff benefits, slashing company allowances for childcare, home internet, and wellness. The same report stated that staff will now have to provide higher ups with a full rundown of their completed work at the end of each week.

Businesses looking for a high-density rack server that can handle very demanding workloads will find HPE's ProLiant DL365 Gen10 Plus could be just the ticket. Supporting AMD's EPYC Gen3 CPUs, this dual-socket server teams up a high core count with a huge memory capacity and packs an impressive range of features into its slim-line 1U chassis.

Support for AMD's EPYC 7763 and 7713 CPUs means the DL365 Gen10 Plus can present 128 physical cores and businesses with (very) deep pockets can specify HPE's 256GB LRDIMMs which allow memory to be pushed to an enormous 8TB. Those on tighter budgets can choose from a wide range of lower-spec EPYC Gen3 CPUs and still go up to 2TB of memory using cheaper 64GB RDIMMs.

Naturally, the 1U high chassis limits storage potential but all base units start with eight SFF hot-swap drive bays which can be boosted to ten by replacing the front media bay with an extra dual-drive cage. There's a storage option for every workload scenario too, as the server supports U.2 or U.3 NVMe SSDs along with standard SAS3 and SATA HDDs and SSDs.

HPE ProLiant DL365 Gen10 Plus review: Security and management

Platform and supply chain security is assured as the server's iLO5 ASIC uses HPE's Silicon Root of Trust fingerprinting for firmware validation and will stop the server booting if any tampering is detected. The Automatic Secure Recovery feature validates the iLO firmware and if the test fails, it automatically flashes it with the resident recovery image and also provides facilities for swapping between the active and redundant ROM and scheduling firmware scan and repair processes.

The iLO5 controller provides a dedicated Gigabit port and offers a smart web console that's crammed with valuable information about the server's status. Along with direct links to HPE support, it allows you to keep a close eye on all critical components and provides extensive fault alerting services with an iLO Advanced licence enabling power metering, virtual media services and full OS remote control.

HPE's OneView app is free if you're only using it to monitor your servers, storage and network environments and we had no problems adding the review server's iLO details to the lab's Hyper-V virtualised 7.2 instance. We found it provides plenty of information about the CPUs, power, and temperatures with tools available for firmware updates, power controls, and remote control sessions.

Enterprises and MSPs will like HPE's free iLO Amplifier Pack which presents a dedicated web management console for discovery and inventory of up to 10,000 Gen8, Gen9, and Gen10 servers and also links up with HPE's InfoSight for iLO4 and iLO5 performance monitoring and predictive fault analysis. And then there's HPE's GreenLake Compute Ops Management cloud service which provides a central web portal for managing geographically distributed IT infrastructures, although bear in mind that any servers added to it cannot also be managed by OneView.

HPE ProLiant DL365 Gen10 Plus review: Build and design

HPE has done a fine job with internal design as for such a well-specified server, it's all remarkably tidy inside with easy access afforded to all areas. The two 16-core 3GHz AMD EPYC 7313 CPUs in our system are topped off with solid passive heatsinks with each socket flanked by 16 DIMM slots and HPE supplied the server with 256GB of DDR4 RDIMM memory spread over sixteen modules.

All cooling is handled efficiently by a bank of seven hot-plug cooling fans lined up between the motherboard and drive backplane. For CPU's up to a 155W TDP, you can use standard fan modules while higher power CPUs, NVMe storage, and 100/200Gb network cards require a full house of performance fans.

There's plenty of room to expand as the server has three risers at the back offering a total of three PCI-E Gen 4 slots. The server doesn't have embedded network ports but you won't need to use any PCI-E slots as the rear OCP 3.0 edge connector supports a big choice of mezzanine cards including quad Gigabit, dual 10/25GbE, single 100GbE, and 200Gb Infiniband.

HPE ProLiant DL365 Gen10 Plus review: Storage

All ProLiant Gen10 server models include HPE's SR100i Gen10 embedded controller chip which for the DL365, supports eight SATA drives and up to two NVMe devices. This is an entry-level solution that provides software-managed RAID0, 1, and 5 for SATA devices and RAID0 and 1 for NVMe SSDs but only for Windows Server operating systems.

Our server was supplied with HPE's Smart Array P408i-p SR Gen10 RAID controller which supports all the usual RAID suspects including RAID6 and comes with 2GB of onboard cache and the optional smart storage energy pack. This model uses a PCI-E slot with its two mini-SAS ports neatly cabled through to our 8-bay backplane - you can save a PCI-E slot by specifying the P408i-a version which snaps into a dedicated slot behind the OCP connector.

You're spoilt for choice with storage devices as you can use SATA and SAS3 HDDs and SSDs or go for a full house of NVMe SSDs. The server supports HPE's Tri-Mode RAID cards which allow you to mix and match NVMe 4, SAS3, and SATA devices but note that you cannot mix U.2 and U.3 NVMe SSDs and must choose the appropriate backplane.

If you want to keep all your front drive bays free for data storage, we suggest HPE's M.2 enablement kit. Costing around £100, it slots into a PCI-E slot, connects to two of the motherboard's embedded SATA ports, and provides two M.2 SSD slots for running your OS.

HPE ProLiant DL365 Gen10 Plus review: Verdict

The ProLiant DL365 Gen10 Plus packs a lot of power into its 1U chassis making it a great choice for rack-dense applications requiring a high core density and a big memory footprint. It's extremely well built and designed with great expansion potential, storage features are extensive and HPE offers a remote management service for every occasion.

HPE ProLiant DL365 Gen10 Plus specifications (as reviewed)

HPE's AMD rack server family has always had a sharp focus on memory-intensive workloads, high-performance computing (HPC), and virtualisation duties, and the ProLiant DL345 Gen10 Plus is no exception. A single-socket design keeps costs down but with a maximum of 64 cores up for grabs and a big 4TB memory capacity, it's capable of offering businesses a low cost per virtual machine (VM).

Prices start at £3,047, which gets you a 16-core 3GHz AMD EPYC 7313P Gen3 CPU, 32GB of DDR4 memory, an 8-bay SFF front drive cage managed by a Smart Array P408i-a SR controller (no storage included), and a single 500W hot-plug PSU. HPE offers a good choice of CPUs with the server currently supporting 19 EPYC Gen3 models up to the 64-core 7763 and 7713P.

Businesses that want plenty of storage capacity on tap will find a lot to like here; the DL345 Gen10 Plus offers an impressive range of storage options as it can handle up to 26 SFF HDDs/SSDs, 12 LFF HDDs or even 24 PCI-E Gen4 NVMe SSDs. HPE's Tri-Mode controller cards are also available for this server and bring U.2 and U.3 NVMe SSDs into play along with support for standard SATA/SAS storage devices.

HPE ProLiant DL345 Gen10 Plus review: Storage decisions

Take your time when choosing a storage configuration as there are a lot of options to consider. Base systems start with either a 4-LFF or 8-SFF front drive box and to increase capacity; you just add more boxes along with the appropriate backplanes and RAID controllers to manage them.

For SFF drives, HPE's standard 8+8+8 drive bay box configuration gives you room for 24 drives at the front which can be pushed to 26 with one dual-bay cage which sits in the centre at the rear. All server models come with an onboard SR100i Smart Array controller which provides software-managed RAID arrays for Windows Server systems only and can handle eight SATA drives or two NVMe SSDs.

The model names of HPE's RAID adapters each tell you what they can do. The E208i-a SR in our system is an essential (no cache), 200-series card supporting eight SAS lanes with internal ports, the 'a' means it's a modular controller that doesn't require a PCI-E slot and is a SmartRAID card as opposed to the MegaRAID-powered MR series.

If you want NVMe SSDs, you must choose between U.2 or U.3 drives as mixing them is not supported, and you'll need to select the correct drive box as well. Four Tri-Mode cards are on offer with the MR416i and MR216i models available as modular cards or PCI-E adapters all supporting a mixture of NVMe 4, SAS3, and SATA devices.

HPE ProLiant DL345 Gen10 Plus review: Design and expansion

Removing the lid reveals an impressively tidy interior with easy access afforded to all key components. The single CPU socket sits in the centre of the motherboard and is flanked on each side by banks of eight DIMM slots supporting up to 1TB using RDIMMs or 4TB with horrendously expensive 256GB LRDIMMs.

Cooling is handled by a bank of six hot-plug fans with lower-specified servers requiring standard models. Go for 24 SFF bays and you'll need the performance models. 180W TDP EPYC CPUs require high-performance fans and if you opt for a 280W TDP 64-core EPYC 7763 CPU - you must specify max-performance fans.

An OCP 3.0 slot is provided underneath the primary PCI-E riser cage and HPE offers a good range of mezzanine cards including quad Gigabit, dual 10GbE, and dual 10/25GbE. There's plenty of room for more expansion as both the primary and secondary risers each present two PCI-E Gen4 x 16 slots.

Along with the dual hot-plug 800W Platinum PSUs in our system, HPE offers Titanium 800W plus 500W and 1600W Platinum versions. They all use the same Flex Slot design so can be easily swapped out for more powerful modules later on if required

HPE ProLiant DL345 Gen10 Plus review: Management

If you thought storage options were extensive, wait until you get to server management as HPE offers a veritable wealth of tools, utilities, and cloud platforms. First up is the server's embedded iLO5 controller which presents a web console packed with status information on all the server's critical components. All ProLiant servers include a Standard licence with an optional Advanced version enabling extra features such as real-time power and temperature metering, full OS remote control and virtual media services.

After adding the server's iLO5 details to the lab's virtualised Hyper-V HPE OneView 7.2 app, we could monitor the health of all our ProLiant servers, gather inventory and apply configuration templates. OneView provides a central console for all your HPE systems and can monitor hypervisors, storage arrays, pools, and switches as well.

We also run HPE's free iLO Amplifier Pack on our Hyper-V server which provides a separate web management console for discovery, inventory, and compliance reporting for up to 10,000 Gen8, Gen9, and Gen10 servers. Its dashboard displays a complete overview of every server and after linking it up with our HPE InfoSight cloud account, we could view details such as critical server health issues, power status and installed hardware from its web portal.

Organisations with geographically distributed on-premises, cloud, and edge compute resources will approve of HPE's GreenLake Compute Ops Management cloud service. From one portal, they can view their entire infrastructure, centrally manage server deployments and automate system lifecycle management and firmware compliance policies.

HPE ProLiant DL345 Gen10 Plus review: Verdict

With AMD EPYC at the helm, HPE's ProLiant DL345 Gen10 Plus 1P rack server offers a core count and memory footprint that are high enough to make it a cost-effective virtualisation alternative to a 2P rack server. Build quality is excellent, it offers a remarkable range of storage configurations with plenty of room to grow with demand and is supported by an abundance of remote management options.

HPE ProLiant DL345 Gen10 Plus specifications (as reviewed)

Lenovo celebrated 30 years of its ThinkSystem x86 servers with an incredible launch of more than 50 new products.

The computing giant revealed a suite of new end-to-end infrastructure services, including its new Lenovo Infrastructure Solutions V3 portfolio.

The introduction celebrates the 30 years of the ThinkSystem x86 server, which is seen as a pivotal breakthrough in enterprise computing technology. One that enabled an open ecosystem for innovation and led the way to a modern information era, according to Lenovo.

With its new Infrastructure Solutions V3 portfolio, the Chinese giant will provide a unique open architecture with advanced management, superior reliability and extended security to help organisation harness a new era of IT. The main aim is to help businesses manage hybrid multi-cloud and edge environments, alongside traditional AI workloads, across retail, manufacturing, the financial sector, healthcare and telecoms industries.

"Lenovo has pioneered infrastructure solutions that have transformed entire industries for more than 30 years," said Kirk Skaugen, the executive vice president of Lenovo. "Building on this legacy, we are now delivering the broadest portfolio advancement in our history, with industry-leading performance, reliability and sustainability, all based on an open architecture that will help customers more easily harness data to solve humanity's greatest challenges."

The V3 portfolio incorporates the next three generations of the Lenovo infrastructure family: ThinkSystem, ThinkAgile and ThinkEdge. These are server and storage systems, all recognised as the industry's most reliable and secure, according to Lenovo. The solutions are supported by next-gen processors - AMD EPYC, Intel Xeon and Arm - in addition to AMD Instinct and Nvidia GPUs.

For simplified edge-to-cloud management, Lenovo also unveiled XClarity One, an open cloud software platform that incorporates TruScale infrastructure as a service (IaaS) with management as a service (MaaS) into a single customer portal.

The Lenovo TruScale IaaS will provide a pay-as-you-go model, eliminating over-provisioning and providing real-time usage cost insight so that customers can also easily customise where management data resides. Whether that's from public to private or across hybrid multi-cloud environments.

The news that Insteon, purveyors of Internet of Things (IoT) smart home products, had shut down in April will have come as an unpleasant surprise to its erstwhile customers. The servers went offline, as did the website’s support pages. Although a group of customers have endeavoured to resurrect the cloud functionality of Insteon devices by buying out the company, this was always seen as highly unlikely.

I’m not surprised there was – albeit temporarily – a casualty like this. Any product that relies on an intimate relationship with a cloud service for it to function is at risk of that service failing or disappearing altogether.

We might wonder how this situation has come about. There is no intrinsic need for a cloud service to be involved with the core operation of an IoT device, but there is an interesting business model at play. For a start, it means the product can just call out to a cloud service somewhere hosted “out there”, with no need for the smarts to be running within the local area network (LAN). This makes deployment and set up easier.

Even better, for manufacturers, is that it gives them one central place to update capabilities and functionality for all their customers. No need to talk to a big bunch of local LAN controllers: just update the cloud server. If you have a local networking box, maybe to allow a protocol transition to Zigbee, then this box can be essentially dumb and closed to the user. And it’s much easier for an app on your phone to communicate with the cloud server as it saves any pesky need for local LAN connection, virtual private network (VPN) tunnels or any other kind of “at home, away from home” configuration hassle.

One more thing: it opens up entirely new revenue streams for manufacturers by forcing the user to buy into a cloud subscription service. The more you pay, the more features you get. If you really want a laugh, go look at the subscription pages for many cloud subscription-based devices, and marvel at the incoherent and confusing tiers and other plans that are foisted upon the customer.

It’s undoubtedly an effective business model, right up to the point where the cloud service disappears. Then you, the customer, are left with junk.

The easy answer is to say, “avoid any product that has any element of cloud functionality”. But that is too simplistic and doesn’t take into account a realistic evaluation of the sturdiness of the company involved. For example, it’s clearly possible for a vendor like Insteon to come off the rails, but the warnings were there – apparently it was trying to structure a buy-out some months ago. The problem is that there is no meaningful route to warn customers of impending doom. As soon as you do, there is a real chance that a small problem cascades into a bigger, company-killing issue.

So, we must do our own realistic risk assessments. It’s possible that Amazon might pull the plug on the entire Alexa infrastructure, but I would suggest that it’s pretty unlikely. Apple might walk away from Siri one day, but it wouldn’t be because the company has run out of money. And you would get months, perhaps years, of warning.

Compare and contrast this level of certainty with that of much smaller tier two, three and four suppliers. While they might well integrate into the top tier platforms such as Azure, they’re still reliant on their own cloud services infrastructures. Consider my home security cameras: I am passionately convinced that these should be wired up using Ethernet cables, preferably with power coming from the POE standard. There should be a local network video recording solution that captures everything, 24/7. And there is no hook into a cloud service.

I can do this because I’m prepared to put industrial-grade networking in place to support the services required to perform a security function. This isn’t quite so appropriate if you want a camera to monitor your bird feeder, to make sure you don’t miss the arrival of a lesser spotted warbler. If so, then a cloud-based solution might be appropriate, but I would certainly only consider one from the top tier companies, and preferably one that had local storage. And which supports standards like ONVIF so it can be hooked up to a NAS/video recording box.

The problem is that the proverbial people on the Clapham Omnibus are bad at doing this assessment, because they simply want something that works. What happens in six months’ time, let along six years, is generally not on their radar.

The failure of Insteon won’t be a sole event. The ongoing supply chain constraints are putting extreme pressure on the smaller vendors. In the past the saying maintained “no one got fired for buying IBM”. There must be a modern equivalent, whether that is Amazon, Apple, Google or another of the big players. Just don’t fall in love with a relatively unknown vendor and put too many capabilities in the one basket. It might not be there tomorrow.

Microsoft's SQL Server is easily one of the world's most popular databases but as with any other solution, it needs to be constantly monitored to stay ahead of performance issues and ensure they don't impact on the delivery of critical services. Database administrators can use the activity monitor and data collector tools built into Microsoft's SQL Server Management Studio (SSMS) but the former isn't overly informative and can be resource hungry while the latter isn't particularly easy to use either.

Businesses are spoilt for choice with third-party database monitoring tools but those that want one that focuses purely on SQL Server without any extra baggage will find SolarWinds SQL Sentry fits the bill nicely. It goes much further than providing full visibility of on-premises database estates; it can keep a close eye on Azure SQL Database and SQL Server Analysis Services plus SQL Server on Amazon RDS and EC2 as well.

Virtualisation gets plenty of attention as SQL Sentry monitors and reports on Windows Hyper-V and VMware systems hosting SQL Server. This allows it to provide a full picture of all databases and associated host systems to help troubleshoot, fix problems and optimise performance.

SolarWinds SQL Sentry review: Deployment

Pricing for SQL Sentry starts at £1,100 exc VAT per year for each monitored instance with volume discounts available. You can also purchase a perpetual license for £2,120 and we were advised that the yearly maintenance renewal fees for this are £465 per instance.

We found SQL Sentry easy enough to deploy and installed it in 30 minutes on a Windows Server 2019 Hyper-V VM. Note that it requires a separately purchased SQL Server database for storing all performance, diagnostics and reporting data but for testing in our lab, we loaded Microsoft's free SQL Server 2019 Express.

SQL Sentry runs a quick on-boarding routine where we pointed it at our SQL Express instance and ran through setting up secure access to its web portal. The SQL Sentry client app is where most of the action takes place and after adding the details of the SQL Server hosts we wanted monitored, it checked the availability of Windows metrics such as CPU, memory, processes and storage activity and added them to the explorer menu in the left pane for easy access.

SolarWinds SQL Sentry review: The SQL Sentry Client

Database monitoring begins immediately and we were bowled over by the amount of information provided by the client app. Our test SQL Server database was hosted on a VMware VM and the client's dashboard view presented no less than 15 graphs.

The left section focuses on the host with real-time graphs showing network, CPU, system memory and disk utilization. To the right are graphs for database activity, lookups, backup throughput, waits, memory usage, database I/O and much more.

It's a busy interface but the upper ribbon menu can be used to pause the auto-refresh and swap from real-time views to specific time periods so you can easily look back and inspect earlier events. Drag the mouse across an area of interest in one graph, and SQL Sentry highlights the relevant areas in all the others for quick event correlation.

The dashboard view is accompanied by tabs across the top for quick access to views such as host processes, disk activity and space usage, the top SQL queries, query plans and their history, SQL Server blocking analysis and deadlocks. We particularly liked the disk activity view as it provides a colour coded row of icons across the top for each database accompanied by active traffic paths so you can see which ones are being used the most.

If you have problems diagnosing complex database deadlocks, you'll love SQL Sentry as it provides an Outlook-style calendar showing when they occurred along with facilities to drill down deeper. Use the smart playback service and you can view the sequence of events for the selected deadlock, the successful and unsuccessful lock requests and code rollbacks along with the affected host, application, database and statement.

SolarWinds SQL Sentry review: The web portal

The SQL Sentry web portal is equally informative and opens with a dashboard overview of all monitored databases and Windows hosts, along with an overall health score and a list of all the latest alerts. Swap to a specific database server and the view changes to show all its alerts graded by severity with options to view them by tag such as CPU, memory and disk or by wait time.

The view provides the same wealth of performance metrics as provided by the client app along with tabs to view the top SQL queries, all blocking activity and deadlocks plus tempDB activity and session collections. The views can be set to custom date ranges and for deadlocks, you get the same detailed information as in the client app although not the playback feature.

The portal can present multiple custom dashboards where you choose which host and database performance metrics you want displayed. Each dashboard can show the latest action in real-time or you can pause the live data feed and swap to a custom date range.

SolarWinds SQL Sentry review: Verdict

Administrators engaged in a daily fire-fight trying to keep their SQL Server databases at peak performance will find SolarWinds SQL Sentry provides everything needed to ensure they run like well-oiled machines. Both the client app and web portal provide a staggering amount of information about host and database performance and the smart deadlock analysis service is a standout feature.

The sheer range of features can present a steep learning curve but SolarWinds provides good online documentation and we found both consoles to be well designed and fairly easy to navigate. It scales easily across businesses of all sizes and SQL Server database estates and for what it offers, is well worth the asking price.

AMD and multi-cloud MSP DoiT International have announced a consultancy collaboration to help Google Cloud customers manage their compute workloads.

The two companies will combine their technical expertise to identify the right workloads for benchmark comparisons on AMD Epyc processors so that Google Cloud customers can optimise cloud spending and management.

Google Cloud has been adopting AMD's third-gen Epyc server processors since they became available last year. The company uses them to power virtual machines and they are said to be good for general-purpose and scale-out workloads, where the workload can move off position and incur more costs.

The collaboration aims to help customers spot where they might find avoidable price changes.

DoiT describes itself as a managed service provider (MSP) that offers consultancy and also products that simplify the automation of cloud workloads. It's a long-term Google Cloud partner that runs international operations across 70 different countries.

Along with AMD, DoiT will provide Google Cloud customers with "industry best" tools, technology and technical advice in a bid to fine-tune compute workloads. Performance and longevity are also key areas of focus, but the main attraction of the deal is that customers will have access to DoiT's advanced technology suite, which includes its FlexSave offerings, for free.

FlexSave is an autonomous service that analyses workloads that aren't already covered by compute commitments or committed use discounts. The idea is about avoiding unforeseen costs on workloads that are not immediately necessary to the business but are still costing the full price.

The collaboration will combine what AMD and DoiT do best, according to DoiT CEO Yoav Toussia-Cohen, which is "delivering tangible, meaningful results for customers".

"For more than a decade we've worked alongside digitally savvy customers of all sizes, all struggling with the fundamental challenge of improving performance without inflating cloud cost," said Toussia-Cohen, CEO of DoiT International.

Alibaba has unveiled a new in-house processor that will be used to power servers in its data centres, as China aims to increase its domestic chip production in the face of a global chip shortage and US sanctions targeting the country.

The server chips, named Yitian 710, are custom-built by the company’s chip development business, named T-Head. It’s powered by 128 Arm cores with a 3.2GHz top clock speed and is the first server processor compatible with the latest Armv9 architecture. It includes 8 DDR5 channels and 96-lane PCIe 5.0, providing high memory and I/O bandwidth.

Alibaba has also developed proprietary servers, called Panjiu, for the next-generation of cloud-native infrastructure. The company said that by separating computing from storage, the servers are optimised for both general-purpose and specialised AI computing, as well as high-performance storage.

The servers have been developed for large-scale data centre deployment through their modular design, and the company expects them to serve a variety of cloud-native workloads such as containerised applications and computed optimised workloads.

"Customizing our own server chips is consistent with our ongoing efforts toward boosting our computing capabilities with better performance and improved energy efficiency," said Jeff Zhang, president of Alibaba Cloud Intelligence and head of Alibaba DAMO Academy. "We plan to use the chips to support current and future businesses across the Alibaba Group ecosystem. We will also offer our clients next-generation computing services powered by the new chip-powered servers in the near future."

Zhang added that, together with Intel, Nvidia, AMD, and Arm, Alibaba will continue to innovate its compute infrastructure and other diverse computing services for global customers.

Alibaba also announced it would open the source code of the XuanTie IP core series, the company’s custom-built processors based on RISC-V instruction-set architecture. Developers will not be able to access the cores’ source code on GitHub and Open Chip Community to build prototype chips of their own.

This coincides with the Chinese government’s “Made in China 2025” initiative of lifting the country’s chip production from less than 10% at the time to 40% in 2020 and 70% in 2025, as reported by NIkkei Asia. The government has invested in the semiconductor industry to try and meet these goals although this appears to be a long way off considering data from IC Insights shows that the country only sourced 16% of its semiconductors domestically.

Being self-reliant on chips is a key issue for the country, especially as it has been hit by a number of US sanctions targeting the tech sector in recent years. In April this year, the US government added seven Chinese supercomputing entities to its Entity List, restricting US exports to them. The US government adds entities to this list as it feels they are working against US national security or foreign policy interests.

Provided by

The popularity of serverless strategies is growing because they provides the opportunity for faster time to market by dynamically and automatically allocating compute and memory based on user requests.

A serverless strategy also provides cost savings through hands-off infrastructure management, which enables organisations to redirect IT budget and development resources from operations to innovation.

In 2019, Deloitte introduced a framework for comparing the total cost of ownership (TCO) for both serverless and server-based applications, factoring in infrastructure, development, and maintenance costs. Read this eBook to learn more about the TCO of serverless and server-based apps.

A hacker group has claimed to have breached the servers of Acer India, with approximately 60GB of sensitive data belonging to several million of the company’s customers being leaked online.

Known as Desorden, the group said it had stolen customer information, corporate data, financial data, and information related to recent company audits, according to a post on a popular hacking forum, seen by Privacy Affairs researchers.

The hackers said that the breach includes data on several million Acer customers, mostly from India. It appears to have taken place on 5 October, as this is the most recent date listed in the leaked databases.

Desorden also said that it will give Acer access to the database to verify the data and prove the breach is real. A sample of the data released for free, which included information on over 10,000 individuals, was found to be accurate and genuine by researchers at Privacy Affairs, who were able to make contact with some of those affected.

The group has said that data belonging to several million more Acer customers will be released for a fee at a later date.

“We have recently detected an isolated attack on our local after-sales service system in India,” an Acer spokesperson told IT Pro. “Upon detection, we immediately initiated our security protocols and conducted a full scan of our systems. We are notifying all potentially affected customers in India.”

The spokesperson added that the incident has been reported to local law enforcement and the Indian Computer Emergency Response Team, and there has been no material impact to the company’s operations and business continuity.

In March this year, Acer fell victim to a $50 million ransomware attack carried out by the notorious ransomware gang REvil. The group announced the Acer breach on its website where it presented images of allegedly stolen files, including financial spreadsheets, bank communications, and bank balances. The breach was believed to be linked to the Microsoft Exchange cyber attack, which was carried out by at least ten hacker groups.

A faulty configuration change has been blamed for taking Facebook, WhatsApp and Instagram offline for more than six hours on Monday night.

The social network's engineering team said that the changes affected the routers that coordinate the platform's network traffic between its data centres. This, they said, caused a "cascading effect" on the way its data centres communicate, bringing all of the company's services to a halt.

"Our services are now back online and we're actively working to fully return them to regular operations," the company said in a blog post. "We want to make clear at this time we believe the root cause of this outage was a faulty configuration change. We also have no evidence that user data was compromised as a result of this downtime."

In order to remedy the issue, Facebook sent engineers to one of its main data centres in California, according to The New York Times, suggesting it couldn't be fixed remotely. It was also reported that the outage prevented staff from accessing company buildings and conference rooms with their badges.

The incident caught the attention of internet giant Cloudflare, which initially assumed something was wrong with its own DNS servers. However, after an investigation, engineers realised something more serious was happening, and reported in a blog that "social media quickly burst into flames."

"Facebook and its affiliated services WhatsApp and Instagram were, in fact, all down," Cloudflare said. "Their DNS names stopped resolving, and their infrastructure IPs were unreachable. It was as if someone had 'pulled the cables' from their data centres all at once and disconnected them from the Internet."

The issues were down to BGP - the Border Gateway Protocol - which is a mechanism that exchanges routing information between autonomous systems on the web. The bigger versions of these make the internet work and have constantly updated lists for the possible routes of traffic, according to Cloudflare.

"The Internet is literally a network of networks, and it's bound together by BGP," the firm said in its blog. "BGP allows one network (say Facebook) to advertise its presence to other networks that form the Internet. As we write Facebook is not advertising its presence, ISPs and other networks can't find Facebook's network and so it is unavailable."

Cisco has announced that it is helping defence departments modernise and simplify their IT infrastructure with its AMD-powered rack servers.

The UCS C4200 Series Rack Server Chassis hosts four UCS C125 M5 Rack Server Nodes in two rack units (2RU) with shared power and cooling. The server nodes are powered by AMD EPYC processors, which boast “the highest core density in the industry”.

This has made it possible to cut down on the number of cables used by defence departments, reducing the number from 200 individual cables by 60% to only 80.

Cisco has managed to merge 20 racks of gear down to a single rack, as well as cut down on power consumption and licensing costs, making running the technology more affordable for the defence departments, which are typically funded by taxpayers.

The new offerings make it easier to manage servers: in a case study, Cisco detailed that “one defence agency deployed the Cisco UCS C-Series to simplify infrastructure management and scaling”.

The unnamed defence department no longer has to manually manage servers on an individual basis, and can now use UCS Manager to orchestrate them “collectively using software-defined service profiles”.

Cisco’s UCS Manager simplifies the deployment of service profiles to both rack and blade servers, with defence departments being able to manage as many as 160 nodes at the same time.

“As a result, the defence agency has streamlined infrastructure management, established greater consistency of server configuration and security, and simplified scaling without the need for downtime,” said Cisco.

The combination of the UCS C4200 Series Rack Server Chassis and UCS C125 M5 Rack Server Nodes is used to benefit “various defence departments in multiple countries”. However, Cisco didn’t specify which exact states are customers.

One of the benefactors could be the US Department of Defense (DoD), which has strong ties with the American tech giant. In July, Cisco launched Webex for defence, an all-in-one collaboration platform specifically made for the military department. Authorised to work with the DoD’s national security systems, the new tool integrates with Cisco’s full Webex portfolio of devices, allowing users to connect securely from phone, desktop, or video.

Provided by

Many CIOs and IT leaders try to extend the lifespan of older servers to reduce costs, but IDC believes that this is expensive and risky in the long run as older servers are less reliable, more costly to maintain, and can have indirect effects such as lower employee productivity.

Server infrastructure generates tangible return to your business, but it can depreciate quickly as each new generation of servers performs much better. To choose the optimal time for replacement, you need to be able to measure the total value of a server by calculating direct and indirect costs.

This IDC opinion piece helps IT leaders fully comprehend the costs of delaying server replacements through interviews and surveys with IT decision makers and practitioners. Download it now to get essential guidance on upgrading.

T-Mobile has confirmed that the records of 47.8 million current, prospective, and former customers have been accessed by hackers, which represents nearly half of the 100 million records that were found for sale online earlier this week.

The company’s preliminary analysis has shown that around 7.8 million current T-Mobile postpaid customer accounts’ information appears in the files, as well as just over 40 million records of former or prospective customers who had previously applied for credit with T-Mobile.

“Some of the data accessed did include customers’ first and last names, date of birth, SSN, and driver’s license/ID information for a subset of current and former postpay customers and prospective T-Mobile customers,” the company confirmed.

It said that, so far, there is “no indication” that the data contained in the stolen files included any customer financial information, credit card information, debit or other payment information. adding that around 850,000 active T-Mobile prepaid customer names, phone numbers, and account PINs were also exposed.

The company has proactively reset all of the PINs on these accounts and will start notifying customers accordingly “right away”.

T-Mobile revealed it is taking “immediate steps” to help protect all individuals who may be at risk from this "cyber attack". It is set to offer customers two years of free identity protection services with McAfee’s ID Theft Protection Service, and recommend its postpaid customers to change their PINs, despite the fact that it has “no knowledge” that any of these PINs were compromised.

It will also offer “an extra step” to customers to protect their mobile accounts and also publish a unique web page later today for “one stop information and solutions” to help customers take steps to further protect themselves.

On 16 August, T-Mobile launched an investigation into a claim on an online forum that suggested the personal data from over 100 million users had been breached. The seller said that the data had come from T-Mobile servers and was asking for six Bitcoin, around $270,000, for a subset of the data containing around 30 million records. A day later, the company confirmed that there had been some “unauthorized access” to its data, but hadn’t determined if any personal customer data had been involved until today.

UPDATE: T-Mobile has confirmed that data belonging to the company may have been “illegally accessed”.

“We have determined that unauthorized access to some T-Mobile data occurred, however we have not yet determined that there is any personal customer data involved,” the company said in a statement to IT Pro. “We are confident that the entry point used to gain access has been closed, and we are continuing our deep technical review of the situation across our systems to identify the nature of any data that was illegally accessed.”

T-Mobile added that the investigation will “take some time” but it is working with the “highest degree of urgency”.

“Until we have completed this assessment we cannot confirm the reported number of records affected or the validity of statements made by others,” it said.

The company also said that once it has a more complete and verified understanding of what occurred, it will then communicate with its customers and stakeholders.

16/08/21: T-Mobile has launched an investigation into a claim on an online forum which suggests that the personal data from over 100 million users have been breached.

The forum post doesn’t explicitly mention the company, but the seller told Motherboard they have obtained data related to over 100 million people and that this data came from T-Mobile servers.

The data reportedly contains social security numbers, driver license information, phone numbers, physical addresses, and unique IMEI numbers. Motherboard saw samples of the data and confirmed they contained accurate information on T-Mobile customers.

On the forum, the seller is asking for six Bitcoin, which is approximately $270,000, for a subset of the data which contains 30 million social security numbers and driver licenses.

"I think they already found out because we lost access to the backdoored servers," the seller told Motherboard, referring to T-Mobile's potential response to the breach.

Despite this, the seller said they had already downloaded the data locally and it is backed up in multiple places.

"We are aware of claims made in an underground forum and have been actively investigating their validity,” T-Mobile said in a statement to IT Pro. “We do not have any additional information to share at this time."

Ilia Kolochenko, Founder of ImmuniWeb and a member of Europol Data Protection Experts Network, said that the price for the records is "very cheap", at just 1 cent per victim. He said the data could be exploited to conduct targeted mobile attacks, social engineering, sophisticated phishing campaigns, or financial fraud.

"From a legal viewpoint, if the information about the breach is confirmed, T-Mobile may face an avalanche of individual and class action lawsuits from the victims, as well as protracted investigations and serious monetary penalties from the states where the victims are based," he said, adding that it would be premature to make a conclusion before T-Mobile makes an official statement on the quantity and nature of the stolen data.

In January this year, T-Mobile suffered a data breach affecting information government agencies considered to be highly sensitive. It affected around 200,000 customers and contained information such as customer phone numbers and the number of lines subscribed to on their account.

Security researchers have discovered a new variant of the Ryuk ransomware that is targeting web servers.

According to a blog post by Marc Elias, a security researcher on the McAfee Advanced Threat Research team, Ryuk ransomware has shifted its attention to web servers since it no longer encrypts the index file but replaces it with the ransom note instead.

Elias said that the Ryuk infection chain usually starts with a spear phishing email that includes a malicious URL or Office document to gain initial entry into victim environments.In certain cases, compromised RDP computers provide the initial access.

In the first scenario, either Trickbot or BazarLoader will be executed and used as a loader malware, offering other actors the opportunity to purchase hacked machines.

Once access to the victim’s machines is acquired by the ransomware actors, a Cobalt Strike beacon is often downloaded in order to obtain users’ credentials and move laterally on the network to take over the domain controllers. Finally, the Ryuk binary is distributed to every machine from the domain controllers.

Elias said that Ryuk copies itself three times in the current directory with different names and launches these new executables with distinct command lines to execute different functionality in each execution.To notify the user about the encryption, Ryuk drops an HTML ransom note in every folder that it encrypts.

“This note is remarkably similar to the note used in other Ryuk variants, with the only difference being the use of a contact button with some instructions to install the Tor Browser,” said Elias.

After file encryption, the ransomware will print 50 copies of the ransom note on the default printer.

Elias said that organizations should be on the lookout for traces and behaviors that correlate to open source pen test tools such as winPEAS, Lazagne, Bloodhound and Sharp Hound, or hacking frameworks like Cobalt Strike, Metasploit, Empire or Covenant, as well as abnormal behavior of non-malicious tools that have a dual use.

“These seemingly legitimate tools (e.g., ADfind, PSExec, PowerShell, etc.) can be used for things like enumeration and execution. Subsequently, be on the lookout for abnormal usage of Windows Management Instrumentation WMIC (T1047),” he said.

Elias added that in the first half of the year, several Ryuk actors have been known to be actively launching new campaigns and targeting organizations all over the world.

“This is the reason we believe the criminals behind Ryuk will continue to develop new features and invent new methods to maximize their profits,” he added.

Microsoft Teams has suffered a global outage for the second time this month that is preventing users from sending messages and logging in to the service.

The tech giant confirmed the issue on Twitter and said it was currently investigating the root cause.

The issue, which is throwing up a 401 error code, is preventing users from accessing the service via the web and blocking messaging systems on the app itself. Some have managed to log in, but Microsoft has warned that they might experience "degraded performance with multiple features".

"We've confirmed that this issue affects users globally," says Microsoft's 365 status account on Twitter. "We're reviewing monitoring telemetry and recent changes to isolate the source of the issue."

The problems began around 10:30am BST, according to DownDetector, with login and server issues the most commonly cited by users. Other Microsoft services, such as Azure and Xbox have so far been unaffected, but with the pandemic still lingering and many still working from home, the outage is causing trouble for many in the UK and around the world - including Doncaster's local planning committee.

"Unfortunately due to global issues with Microsoft Teams today's scheduled Planning Committee is cancelled," the organisation said on Twitter. "The meeting will be rearranged in due course, we will share details of the new date once it becomes available."

Since, Microsoft has said that it's identified the root cause of the issue and "performed mitigation actions".

"We’re seeing signs of recovery and will continue to monitor the service," the company added. "Users still experiencing impact should restart their clients to expedite recovery. Further details can be found under TM252802.

This is the second outage Teams has suffered in April and the third its experienced in the past two months. In March, users lost four hours of usage after an authentication change knocked out access to the comms platform. That coincided with an Azure Active Directory outage that impacted Office web apps, Exchange Online, SharePoint Online and other Microsoft services.

IT Pro has approached Microsoft for details and will update this article when more information is released.

An HTTP Error 400, often referred to as the "Bad Request" error, occurs when a website server is unable to process a request due to issues such as an incorrect URL or invalid request data. Unlike server-side errors like HTTP 502 or 503, a 400 error typically originates from the user’s end. This means the fix is often straightforward, such as correcting a URL or clearing out old cookies.

Depending on the browser you are using, an HTTP Error 400 can result in a blank page, or a page showing a more generic, user-friendly message, as is the case with Opera and Chrome.

However, there are a bunch of other messages you might encounter that all signal an HTTP Error 400, including:

• Bad Request - Invalid URL
• Bad Request: Error 400
• HTTP Error 400 - Bad Request
• HTTP Error 400. The request hostname is invalid
• Bad Request. Your browser sent a request that this server could not understand

Website owners are able to customise the page that an HTTP Error 400 will display to visitors. An example of this is when web servers run Microsoft Internet Information Services (IIS), which results in a page breaking down the error in more detail, such as “400.3: Invalid If Header”, “400.2: Invalid Depth Header”, “400.1: Invalid Destination Header”, and more.

What causes an HTTP Error 400?

Several issues can trigger an HTTP Error 400, which generally stem from user input or browser data. Here are five of the most common causes:

1. Bad URL Syntax

The most frequent reason for an HTTP Error 400 is a misformatted URL. This could be the result of typing errors, such as including special characters that aren’t allowed in URLs (e.g., using a backslash instead of a forward slash). 

You can trigger an HTTP Error 400 manually, usually by typing "/%" after the URL.

For example, the following URL will send you to a valid page, the page to sign up to our daily newsletter:

'https://www.itpro.com/newsletter-signup'

However, if you were to type this URL in manually and type "/%" at the end of the URL, the request will result in an HTTP Error 400 page - as the server has received a 'Bad Request'.

2. Invalid Cookies

Sometimes, outdated or corrupt cookies can trigger an HTTP 400 error. Cookies store website data, including authentication information, but if they become invalid (e.g., due to session expiration), they can cause the server to reject your request.

3. Incorrect file size

In certain cases, an HTTP Error 400 may occur if the file you're trying to upload exceeds the server's size limit. Many web servers impose restrictions on file uploads to ensure performance, and if your file is too large, the request will fail with a 400 error. For example, WordPress sites often limit upload sizes to between 4MB and 128MB.

4. Unsynchronized DNS Cache

A 400 error may result from outdated or mismatched DNS information stored locally in your system's cache. DNS caches store IP address data associated with domain names, but if this information becomes out of sync, your browser may send incorrect data to the server, leading to a bad request error.

5. Server error

While rare, an HTTP Error 400 can also occur due to issues on the server's end. Server misconfigurations or system overloads may result in the server misinterpreting requests as invalid. If you're seeing the error across multiple browsers and devices, this could be the case.

How to fix a 400 Bad Request

Despite being frustrating, resolving a 400 error usually takes just a few simple steps. Here are some common fixes to get your browsing back on track:

When the Problem is Server-Side

If you’ve followed all the steps and the HTTP Error 400 persists, the issue may be on the website’s server. Unfortunately, in such cases, there’s little you can do other than wait for the site administrators to fix the problem. Reporting the issue to the website owner can help speed up the resolution, and websites like Down Detector can provide information on any widespread outages affecting the site.

Although 400 errors can be perplexing, they often have simple causes like a mistyped URL or outdated browser data. By following these troubleshooting steps, you should be able to quickly identify and resolve the issue in most cases.

How to Prevent Future HTTP 400 Errors

While HTTP 400 errors are often easy to fix, you can take proactive steps to reduce the likelihood of encountering them in the future. Here are some practical tips to help keep your browsing experience smooth:

Microsoft released three new patches for its Exchange Server software on Tuesday after the National Security Agency (NSA) alerted the company to a fresh batch of critical vulnerabilities.

The new fixes are for three versions of Exchange Server - 2013, 2016 and 2019 - and the flaws are said to be different vulnerabilities to the ones disclosed in March. However, US agencies continue to find and remove vulnerabilities in their systems a month after the previous flaws were first discovered.

In response to the release of new fixes, the White House ordered all its agencies to install them, warning that the vulnerabilities "pose an unacceptable risk" to Federal operations.

Microsoft's Exchange Server email and calendar software is mostly used in on-premise data centres. The popularity of the system was highlighted by the number of reported breaches the followed the discovery of the initial flaws.

"Microsoft released a set of Exchange patches today that are critical," a White House statement read. "We urge all owners and operators of Microsoft Exchange Servers to apply these latest patches immediately. The US government will lead by example - we are requiring all agencies to immediately patch their Exchange servers, as well."

Exchange Server vulnerabilities have caused issues for a number of organisations around the world, with many servers having already been breached and still vulnerable via embedded back doors. China state-sponsored hacking group Hafnium was spotted by Microsoft using the vulnerability to break into Exchange Servers to view or steal contents.

These vulnerabilities were patched by Microsoft, but backdoors embedded in breached servers were not closed. Within a few days, other hacking groups began hitting compromised servers with the same flaws to deploy ransomware.

As a result, a US court has had to authorise an FBI operation to "copy and remove" backdoors from hundreds of Exchange Servers. The Justice Department said the operation was "successful", but it only removed backdoors and did not patch the vulnerabilities exploited by the hackers or remove any malware that may have been left behind.

The founder of OVH, which owns the French data centre that perished in a fire last week, has said the incident highlights a need for the data centre industry to offer backups as a standard for all customers.

Octave Klaba said his company will start providing secure backups for its data centre customers by default rather than as an additional paid service.

Klaba said that the fire at OVH's data centre site in Strasbourg earlier this month, that destroyed one building and partially damaged another, should serve as an industry-wide wakeup call.

"This incident will change our way of delivering these services, but, also, I believe it will change the industry, which will increase the securities of backups by default, without any payment," Kalaba said in a video. "This will be our strategy, our answer to this incident."

Several OVH customers affected by the fire were unable to bring their applications back online due to a lack of backups. Klaba suggested there was some confusion over service terms and that some of its customers hadn't fully understood what they had brought from OVH. The French cloud firm did offer 500GB of free backup storage with every dedicated server and customers could pay to ramp that up to 10TB.

"It seems that globally, the customers... understand what we are delivering, but some customers, they don't understand what exactly they bought," Klaba said in the video.

The CEO added that he would post another video on Friday 19 March with an update on the ongoing investigation into the cause of the fire. Thermal images used by firefighters suggested the building's uninterruptible power supply (UPS) system was a possible cause. Klaba said that investigators had taken the UPS units and all its batteries and fuses - along with video footage - for analysis.

AMD has unveiled its third-generation Epyc Milan data centre CPUs, which is claims are the "world's fastest" server processors.

The hardware, which is already available, is the first set of AMD chips for the server market to be built on the company's powerful 7-nanometer Zen 3 architecture.

AMD claims Epyc Milan produces twice the power of Intel's rival Xeon Cascade Lake Refresh chips in HPC, cloud and enterprise workloads, and claims it offers a better price-to-performance ratio.

AMD's confidence comes from as a result of its Zen 3 architecture, which is a microarchitecture that has brought similar benefits with AMD's Ryzen 500 series of chips, which have become increasingly popular in the desktop PC space.

"With the launch of our 3rd Gen AMD Epyc processors, we are incredibly excited to deliver the fastest server CPU in the world. These processors extend our data centre leadership and help customers solve today's most complex IT challenges, while substantially growing our ecosystem," said Forrest Norrod, senior vice president and general manager, Data Center and Embedded Solutions Business Group.

"We not only double the performance over the competition in HPC, cloud and enterprise workloads with our newest server CPUs but together with the AMD Instinct GPUs, we are breaking the exascale barrier in supercomputing and helping to tackle problems that have previously been beyond humanity's reach."

'Zen 3' refers to cores of which there are 64 within the 7003 series that is said to deliver up to twice the performance for HPC workloads compared to the competition. For cloud providers that need more computing power and security, the 7003 series has the "highest core density", according to AMD, and also includes AMD Infinity Guard protection.

Already, the AMD Epyc Milan series is being used by some of the biggest names in tech in a variety of ways. AWS will add it to its core Amazon EC2 instance families later this year, and it will also feature in Dell's new PowerEdge XE8545 server. Google Cloud will embed the EPYC 7003 series processors in a new compute optimised virtual machine and Microsoft has announced multiple virtual machine offerings that will be powered by the 7003 series.

Other uses cases include Lenovo, Oracle, HPE, VMware, Tencent and many more.

An investigation is underway to determine the cause of a data centre fire in the French city of Strasbourg that resulted in the loss of data and service outages across Europe.

The fire tore through one building and partially damaged another on a site operated by French cloud firm OVH.

The cause of the fire is yet to be established, but an investigation has been launched.

The incident was first reported just before half three on 10 March (UTC) and although firefighters responded almost immediately they were unable to stop a blaze inside OVH's SBG2 building. Four rooms inside SBG1 were also destroyed in the fire.

Two other data centres owned by OVH were not affected by the fire, but the company has had to switch off every one of its servers.

"If your production is in Strasbourg, we recommend to activate your Disaster Recovery Plan," the firm warned. "All our teams are fully mobilised along with the firefighters. We will keep you updated as more information becomes available."

OVH has reserved servers at other cites in Roubaix and Gravelines that are ready to support the majority of its affected customers, and an additional 10,000 new servers are already in production. The company added that an official customer announcement and FAQ will be posted on its website soon.

OVH also said it is currently working on a plan to relaunch the two unaffected data centres in Strasbourg, and potentially the partially damaged one, once the site is secured and reconnected to power.

"We ask that our customers exercise caution around the emails they receive: in times of crisis, it is common for malicious activity (phishing, spam, etc.) to increase," OVH warned. "It is more important than ever to stay alert."

OVH has become a popular cloud provider in recent times, with 27 data centres in operation, mainly across Europe. The firm is also set for IPO later in the year.

A number of its customers have taken to Twitter to report downtimes due to the fire. These include digital asset management library Piwigo, gaming cite Rust, and reportedly the French Government.

A tweet from Russia's media watchdog, Roskomnadzor, suggested the outage also took down Google services, including YouTube, in Russia. However, a Google investigation later found that this was due to an upstream network issue and was entirely unrelated to the OVH fire.

Servers come and go but HPE’s ProLiant MicroServer family has stood the test of time. We’ve had them all on the bench over the years; IT Pro reviewed the original N36L model way back in 2011, we wait three years for the slinky silver Gen8, and we were equally impressed with the Gen10 version when it finally made its debut in 2018.

The latest MicroServer Gen10 Plus targets precisely the same market as its predecessors by offering SMBs an affordable entry-level server capable of running a wide range of business apps. This is HPE’s smallest MicroServer to date, but this compactness doesn’t come at the cost of features - there’s a lot going on inside.

It’s all change in the processing department; AMD has had its marching orders, and the Gen10 Plus instead offers a choice of two Intel CPUs. On test, we have the base model with a dual-core 3.8GHz Pentium Gold G5420, but those with heavier workloads and a hankering for VMware virtualization services can go for a quad-core 3.4GHz Xeon E-2224.

There are no compromises with storage either, as the Gen10 Plus also offers four LFF SATA drive bays and any can be converted to use SFF SSDs with enablement kits costing around £10 each. The older MicroServer Gen10 lacked any remote management features, but the Gen10 Plus remedies this, courtesy of an iLO5 enablement kit which costs about £31.

HPE ProLiant MicroServer Gen10 Plus review: Chassis design and storage

Placing the Gen10 Plus next to a Gen10 model shows just how small it is. The Gen10 Plus has a slightly larger footprint but standing at only 119mm, it’s half the height.

HPE has achieved this shortening by changing the drive bay arrangement from four vertically inline to two groups stacked horizontally. The bays are carrier-free and to install a drive, you fit mounting screws on each corner and slide it in until the tab at the front locks – all the screws are provided in the panel below.

On the other hand, there’s no room for an internal optical drive and the power supply goes from internal to external, complete with a chunky 180W power brick. Expansion potential has also been reduced; the Gen10 Plus has one PCI-E Gen 3 x16 slot as opposed to the two in its predecessor.

Basic RAID is provided by the embedded Smart Array S100i SR Gen10 controller, which supports software-managed stripes and mirrors. You can upgrade to fast SAS3 storage as the Gen10 Plus supports HPE’s Smart Array E208i-p SR Gen10 PCI-E adapter which also brings RAID5 arrays into the storage equation.

HPE ProLiant MicroServer Gen10 Plus review: Maintenance and expansion

The chassis cover is retained with two thumbscrews; undo these and it can be removed by sliding it backwards. The front cover has an internal locking tab on each side that can only be accessed with the lid off, so if you use a padlock or Kensington lock, the drive bays behind it can’t be interfered with.

Removing the motherboard is a piece of cake, too - just undo two screws and slide the entire tray out the rear. The only cables to worry about are the SATA and power cables for the drive backplane, which are easily unplugged when the board is pulled out.

Along with quad Gigabit ports, the motherboard has two DIMM slots and both CPU options allow memory to be expanded to a maximum of 32GB. The PCI-E slot is fully accessible with the motherboard pulled back and there was enough room for us to fit a 167mm long Emulex dual-port 10GbE adapter card.

Cooling arrangements are clever, too. The CPU is fitted with a large passive heatsink, and two copper heat-pipes are routed from this to a vertical radiator array that sits in front of the single 10cm diameter system fan at the back. It clearly works well: using the SPLnFFT iOS app on an iPad, we measured a zephyr-like 35.9dB at one metre in front, rising to no more than 37.5dB during prolonged periods of disk activity.

HPE ProLiant MicroServer Gen10 Plus review: System management

The motherboard already has HPE’s iLO5 remote management controller chip onboard and to use it, you’ll need the aforementioned iLO enablement kit (part no. P13788-B21). It’s a great upgrade for the price as it fits in a dedicated bay above the PCI-E slot, provides its own network port and delivers many of the management features found in HPE’s high-end enterprise servers.

The iLO5 web console is packed with useful information about critical components and their status, along with 3D thermal graphs showing temperature ranges throughout the chassis. One missing feature is the power monitoring graph, which is absent due to the server using a ‘dumb’ external PSU.

It’s even better value than it first appears, too - the upgrade incorporates an iLO 5 Essentials license which activates full OS remote control. The lack of optical drive won’t be an issue either as the license also enables virtual media services so you can map a local drive to the server.

Accessed from the server’s boot menu, HPE’s Intelligent Provisioning feature makes light work of OS deployment. We selected this menu option during boot-up, chose our OS from its list, pointed it at the virtual ISO drive we’d mapped to the server and left it to load Windows Server 2019 in 30 minutes.

HPE ProLiant MicroServer Gen10 Plus review: Verdict

With remote working becoming the new normal, many businesses will be looking for an affordable entry-level server that can comfortably run a range of services such as on-premises apps, virtualization or NAS file sharing. The ProLiant MicroServer Gen10 Plus looks an ideal candidate as it takes all the goodness of the Gen10 model, adds powerful Intel processing and packs it all into a chassis that’s half the size.

You will need to factor in the cost of storage as this isn’t included, but a price tag of only £395 for the base system on review is excellent value. The Gen10 Plus is a great choice for remote sites and SOHO deployments as it’s extremely quiet, and the iLO enablement kit option brings classy remote management into play.

HPE ProLiant MicroServer Gen10 Plus specifications

With prices starting at just £538, the PowerEdge T140 is a tempting proposition for growing SMBs looking for their first purpose-built server. This sturdy little tower comes with support for Intel’s powerful Xeon E-2200 CPUs, flexible storage options, and room to expand with demand.

There’s a good range of core configurations on offer to suit almost any workload, too. The Basic entry-level system sports a 3.6GHz Core i3-9100 CPU and 8GB of DDR4, while the Standard model starts at £832 and offers a larger CPU choice, with all Xeon E-2100 and E-2200 models available.

Coming in at £2,033, our review system is clearly a lot more expensive, but it’s powerful enough for the most demanding of roles. We selected a 3.4GHz quad-core Xeon E-2224 CPU, and partnered it with a healthy 32GB of DDR4 RAM: for even more power, you can step up to a six-core HT-enabled Xeon E-2236 for an extra £78, and push the memory up to the supported maximum of 64GB.

Whichever spec you choose, the T140 looks stylish thanks to Dell EMC’s classic honeycomb front cover. In this instance, though, there’s nothing to see behind it, as the server doesn’t support hot-swap hard disks. Instead, storage resides in a cage behind the side panel, which will take up to four cabled LFF hard disks. Installing a quartet of 2TB SATA models was a quick procedure, thanks to the simple-swap, tool-free carriers that slide out of the internal cage. Access is easy too, as the side panel opens up in seconds via a sturdy latching lever. This doesn’t have an integrated key lock, but you can padlock it to block access to the server’s innards.

Standard RAID is provided by the embedded PERC S140 controller, which handles software-managed mirrors, stripes, and RAID5 arrays for SATA drives. For our system, we added the PERC H330 card, which adds support for hardware-managed RAID arrays, plus Nearline-SAS and standard 12Gbits/sec SAS drives.

There’s plenty of room for further expansion too, thanks to the server’s four PCIe slots. We filled one of these with Dell EMC’s BOSS (boot-optimised storage solution) card: for £379, this provides a mirrored pair of 240GB M.2 SSDs for your OS to reside on, leaving the main bays dedicated to data storage.

Internally, the T140 is very well designed, with cable clutter kept to a minimum and easy access to all the key components. Cooling is handled by an active CPU heatsink, plus a discreet 9cm chassis fan at the rear: our iPad recorded a low noise level of 38.5dB from a metre away, so the T140 won’t be a distraction in small offices.

The cherry on the cake is remote management. The PowerEdge T140 is blessed with a full-strength iDRAC9 controller and a dedicated Gigabit Ethernet port, exactly as found on much pricier servers. This presents a slick web console providing a wealth of data on all hardware components, system status, power usage, and cooling efficiency. Along with full firmware and hardware inventory functions, it provides direct access to the BIOS and storage controllers for remote configuration.

Three licensing options are offered for iDRAC9 and we selected the Enterprise level, which also activates full OS remote control and virtual media services. These are very handy for installing an operating system: after mounting a Windows Server 2019 ISO image as a virtual optical drive, we were able to use the embedded Lifecycle Controller to load the OS and all necessary drivers in 30 minutes flat.

The PowerEdge T140 is an ideal candidate for smaller offices that want an affordable on-site server with plenty of room to expand. It packs a powerful hardware package into a space-saving chassis, can be easily customised to suit all budgets, and includes the best remote management features around.

Dell EMC PowerEdge T140 specifications

HPE has secured a deal to build a supercomputer in the Czech Republic in 2021, what is considered to be the fastest of its kind in the country.

The development is part of the EuroHPC Joint Undertaking, an initiative between the EU and the tech industry to coordinate and combine resources to develop world-class exascale supercomputers within the continent.

The new installation has the working title "Euro_IT4", referencing the IT4Innovations National Supercomputing Center, an R&D facility in the Czech Republic, where it will be housed.

HPE will power the project with its Apollo 2000 and Apollo 6500 systems, which are purpose-built to support high-performance computing (HPC) workloads, such as modeling and simulation, with AI and other data-intensive applications.

IT4Innovations has said it plans to use the system to boost weather forecasting, advancing drug discovery – including a cure for COVID-19 – and to develop greener and sustainable infrastructure.

"HPE is uniquely positioned to provide the complete infrastructure and services that the next era of supercomputing demands, and together we are delivering one of the most powerful supercomputers in Europe that will benefit science, industry and society as a whole," said Vit Vondrak, Director at IT4Innovations.

The Euro_IT4 will feature over 500 Nvidia A100 Tensor core GPUs and network for targeted AI performance. Once built, it's anticipated it will sit somewhere between the 20th and 50th fastest supercomputers in the world, with an anticipated peak performance of 15.2 petaflops per second.

Nvidia itself has recently announced plans to build the UK's most powerful supercomputer as part of its acquisition of Arm. The 'Cambridge-1' supercomputer will also be aimed at medical research and would hypothetically rank as the 29th most powerful supercomputer in the global TOP500 list.

AMD's 'Hawk' supercomputer in Stuttgart, Germany, is thought to be the fastest general-purpose system for scientific and industrial computing in Europe. That installation consists of 44 racks and 5,600 computer nodes, providing around 25 petaflops per second of compute power.

That, however, pales in comparison to the world's most powerful supercomputer, the 'Fugaku', which is the number one listed by the TOP500 benchmarking index. Powered by Arm processors, the Fugaku reached 415.5 petaflops per second.

In a large business – particularly a technology focused one – there will likely be an entire IT department focused on configuring the right server solutions for the organisation.

Small and medium businesses (SMBs) will have fewer IT resources and potentially different demands from larger enterprises, and therefore may need more of a steer from external advisers. So how do they ensure that they are getting a server solution that will suit their needs?

The right fit

The first step towards acquiring the server that is right for your SMB is to understand and define the functions that you need it to fulfil for your company. From email to private cloud and everything in between, no two organisations’ server needs are alike, and those needs will define the specifications of your new server.

For instance, Scan’s system integration brand 3XS offers both hardware and software consultancy in order to deliver an IT solution that is optimised for your business requirements and doesn’t needlessly exceed them.

3XS offers a series of preconfigured servers aimed at SMBs on a simple scale of increasing capacity and performance. For companies more concerned with results than the intricacies of a bespoke build, a preconfigured 3XS server is an affordable and hassle-free option. 3XS will consult with you and recommend the server that will best fulfil your needs, and with next day delivery can have you up and running quickly and easily. It goes without saying that, in an IT emergency, that speed of rollout can be invaluable for your business.

3XS’s pre-configured servers are powered by AMD EPYC™ processors, providing an affordable yet powerful and versatile CPU solution, offering unparalleled core density and state-of-the-art security features including AMD Infinity Guard. This unique security feature enables SMBs to feel confident in the knowledge that the risk of attacks are minimised when data is processed, which could be critical when data is your biggest business asset. 3XS servers powered by AMD EPYC™ deliver on the promise of helping lower TCO and enable organisations, no matter the size, to build infrastructure for key workloads without feature compromise.

Services

Identifying the server that best meets your SMB’s needs is only one piece of the puzzle. You want to make sure that you are working with a provider that will continue to support you beyond your purchase – both in the immediate future as you get your server up-and-running and further down the line when those vital workloads start to stack up.

As already discussed, every SMB’s needs are different. You may be in the lucky situation where you or one of your staff is suitably tech-savvy, in which case it may be simple enough to receive your new server from the courier and get it set up and working all on your own. If that is not the case, it is important to know that your provider can guide you through this process, making sure that it’s configured correctly in terms of connectivity, security and access control.

3XS offers a free, value-added virtual setup service to guide you through getting your server up and running, coordinating with you and the courier to find the ideal time slot after your server has been delivered in which to remotely access your system and lead you through the process of getting things set up. The process is inclusive, so that the customer can feel involved and in control of the proceedings, and will be in a position to ask questions about the server setup, filling in any gaps in their technical knowledge along the way.

If virtual setup isn’t enough, you may need to arrange for more hands-on assistance in getting your systems ready. 3XS, for instance, offers a consultancy and install service for an additional cost, which is ideal if you need an expert on-hand to ensure everything is configured the way it should be.

Beyond setup, it is wise to consider the long-term support that your provider offers. Servers will come with a warranty as standard, but how do they compare between providers? 3XS offers a three-year warranty, and will send an engineer out to complete any repairs in the first year. Warranty add-ons are often available if you feel your organisation’s systems require additional protection.

For an SMB, the availability of day-to-day support is an important factor in choosing a server provider. A provider may offer a helpline to troubleshoot any issues, but check what hours they are available. If the lines close at 5.30pm and you experience a serious problem at 5.35pm, you may have to spend a sleepless night waiting for the lines to open again at 9.30am. 3XS’s technical support team is available seven days a week and until 10pm on weeknights.

The majority of problems that 3XS support deals with are simple configuration issues which can be resolved quickly and easily by remotely accessing the user’s system. A good provider acknowledges its role in part of your wider IT team and will be happy to offer advice even when it strays beyond its official remit.

The needs of SMBs are variable. It’s important to identify the server that will help you best achieve your needs, but it’s equally vital to find a provider who you can work with to keep your systems running smoothly and effectively through installation and beyond.

Scan is an Elite Level member of the AMD EPYC Partner Program, awarded in recognition of Scan’s ability and competency to integrate AMD EPYC servers.

Learn more about Scan’s 3XS SMB Infrastructure servers powered by AMD EPYC™

Lenovo announced the launch of two new two-socket servers, the ThinkSystem SR645 and SR665, expanding the company’s portfolio of AMD's EPYC processor-powered data center servers and marking its entry into the dual-socket AMD EPYC solution market.

The launch of these new servers follows last year’s release of the ThinkSystem SR635 and SR655 single-socket servers.

The ThinkSystem SR645 and SR665 two-socket server platforms come equipped with 128 central processing unit (CPU) cores per system and up to two 64-core AMD EPYC 7002 processors.

“Our new Lenovo ThinkSystem servers are designed for workloads such as in-memory databases, advanced analytics, virtualization, and AI,” noted Kamran Amini, VP and general manager of server, storage and software defined infrastructure at Lenovo Data Center Group.

“With the exceptional power, speed, and onboard storage of these new servers, our customers have the ability to handle the increasing data requirements of today’s workloads with the scalability to grow with their business.”

Lenovo created these new servers with an eye on the financial, retail and manufacturing industries, which require faster transaction processing and improved data analytics. These industries also require optimal grid-computing capacity without compromising the efficiency or the total cost of ownership, these new servers aim to deliver in these areas too.

With the addition of the new ThinkSystem SR645 and SR665 servers, Lenovo can help its customers in these industries ramp up higher-performance workloads to improve efficiency.

With their higher core counts, customers can scale down the number of servers they need, making them ideal for areas where rack space is limited. The financial industry, where servers are typically housed in tight quarters near financial exchanges, is a great example where these high-performing servers will work best.

“Lenovo’s endless pursuit of innovations to accelerate our customers’ intelligent transformation has been paramount in our rise to become one of the fastest-growing data center OEMs in the world,” shared Doug Fisher, Lenovo Data Center Group COO, and SVP of business units.

Fisher continued, “Today Lenovo is expanding our AMD relationship with new, fully optimized solutions to help our joint customers address complex and data-intensive workloads, enabling them to do more with less while still providing uncompromised end-to-end security.”

The new Lenovo ThinkSystem SR665 and SR645 solutions are now available through Lenovo sales reps and partners worldwide.

Dell EMC has taken the wraps off PowerStore, a new line of mid-range storage arrays that bring Dell and EMC technologies together in a single appliance, on what would have been the first day of parent company Dell Technologies’ annual conference.

Speaking to journalists ahead of the launch, Travis Vigil, SVP of product management at Dell EMC, said it was the first new product introduced since the 2016 merger between Dell and EMC that used expertise from across both sides of the storage and server business, as well as from other arms – notably VMware.

Caitlin Gordon, VP of marketing at Dell EMC Storage, added that the company has been working on this project for “a number of years”

Explaining the development of PowerStore, Gordon pointed to the fact that data has never been more valuable for businesses than it is today, but noted that it’s also incredibly diverse and difficult to manage. At the same time, organisations are also under pressure to carry out digital transformation, which IT is expected to support.

“What we found in our conversations with customers over the last number of years is they felt like their infrastructure investments require them to prioritise either the needs of their data, or the needs of their operating model in their operations,” she said.

Upon realising there was nothing in the Dell Technologies portfolio – nor, the company claims, anything in the market more generally – that met that dual need in a single system, the decision was taken to build a new product “from the ground up”.

Scalable, programmable, autonomous

PowerStore is an Active-Active HA dual node appliance, with end-to-end NVMe and the ability to support either NVMe-based flash or dual-ported Optane storage class memory (SCM) drives from Intel. This, Dell claims, makes it 7x faster with 3x lower latency than its previous lead mid-range all-flash product.

The company also says that PowerStore can support any workload, traditional or modern, including containers, files, and virtualised or physical apps and databases. It also offers the ability to scale up and scale out up to 2.8 petabytes effective and 11.3 petabytes effective per cluster respectively, as well as having always-on inline deduplication.

It also has built-in machine learning to help optimise system performance, cloudIQ storage monitoring software and is programmable, allowing administrators to treat the infrastructure “as code”.

PowerStore is available immediately.

AMD has launched three new second-gen Epyc server processors which it claims can deliver up to 50% lower cost of ownership compared to rival Intel Xeon processors.

The chips are part of AMD's attempt to top the server chip market, which has long been dominated by Intel.

The latest additions to the Epyc family include eight-core, 16-core and 24-core processors that specifically target database, commercial high-performance computing and hyper-converged infrastructure workloads.

AMD says its second-gen Epyc processors have proven fairly successful in the enterprise, cloud and high-performance computing markets.

"The enterprise is an important area for us," said Dan McNamara, SVP and GM of AMD's server business unit. "We're really going after that with this new launch, with the per-core advantage and per-CPU dollar advantage."

The three new processors are the eight-core Epyc 7F32, the 16-core Epyc 7F52 and the top-end 24-core Epyc 7F72. Each comes with 500MHz of additional base frequency and large amounts of cache memory.

According to AMD, the 7Fx2 processors provide new performance capabilities for workloads at the centre of the enterprise markets. This includes databases with up to 17% higher SQL server performance compared to the competition.

The new processors also offer market-leading per-core performance, according to AMD, thanks to a balanced architecture that combines high-performance Zen 2 cores, innovations in design and the AMD Infinity architecture.

According to the company, its ecosystem of Epyc users continues to grow. This includes hardware makers, cloud providers and software developers, such as Dell, HPE, IBM and more.

"We really feel like not only is the product coming to life across these different segments, but the ecosystem is really starting to grow heavily for us," McNamara added.

Over 350,000 of all Microsoft Exchange servers have not been patched against the CVE-2020-0688 post-auth remote code execution vulnerability impacting all supported Microsoft Exchange Server versions.

A patch arrived in Microsoft's February 11 patch, but few organizations have actually moved forward with applying it to their servers.

Microsoft has encouraged admins to apply the patch as soon possible, tagging it with an "Exploitation More Likely" exploitability index assessment, hinting that the vulnerability may be an attractive target for attackers. The company also said that it anticipates future attacks on the remote code execution vulnerability, making the recently released patch even more important.

Attacks on vulnerable Exchange mail servers began in February. These attacks followed the release of a technical report that detailed how the bug worked. This report was then followed by multiple proof-of-concept exploits, along with a Metasploit module.

Now, nearly two months later, Rapid7 researchers used its Project Sonar to scan the internet and identified at least 82.5% of the 433,464 Exchange servers in the scan were vulnerable to CVE-2020-0688.

To make matters worse, many of the servers tagged by Rapid7 as being safe against attacks may also be vulnerable because many updates to the server did not include an update to the build numbers.

"There are two important efforts that Exchange Administrators and infosec teams need to undertake: verifying deployment of the update and checking for signs of compromise," Rapid7 Labs senior manager Tom Sellers explained.

Compromised user accounts and accounts that have been used in attacks against Exchange servers can be discovered by checking Windows Event and IIS logs for portions of encoded payloads, including either the "Invalid viewstate" text or the __VIEWSTATE and __VIEWSTATEGENERATOR string for requests to a path under /ecp.

Microsoft says that because there are no mitigating factors for this vulnerability, patching your servers before attackers locate them and compromise your entire network is of the utmost importance.

Targeting diverse enterprise workloads such as virtualisation, HCI, big databases and VDI, Dell’s PowerEdge R6515 delivers a monumental core density in the slimmest of rack server packages. Powered by AMD’s Gen2 EPYC 7002 CPUs, our review system was supplied with a 32-core EPYC 7502P but the server is also available with a big choice of other models including four of AMD’s mighty 64-core chips.

The new EPYCs allow the R6515 to deliver a 100% increase in core density over the first generation R6415. The top memory capacity of 2TB and 8 integral channels remain the same, but the EPYC 7002 CPUs boost memory speed from 2,666MHz to 3,200MHz and deliver support for PCI-E 4.0.

The bottom line is the R6515 delivers the processing power and I/O capacity of a 2P server but in a more compact and rack-friendly 1P package. This isn’t to the detriment of other features though, as its nifty internal design allows the server to offer impressive storage credentials and expansion potential.

Dell PowerEdge R6515 review: Design and build quality

Lifting the lid reveals a busy but well-designed interior affording easy access for maintenance maneuvers. The layout is pretty much the same as the R6415 with the EPYC CPU sitting centrally on the motherboard, mounted by a solid passive heatsink and flanked on each side by 8 DIMM slots.

The 7502P CPU in our system has a base speed of 2.5GHz and a top boost frequency of 3.35GHz. It delivers 32 physical and 64 logical cores and sports a meaty 128MB L3 cache.

The EPYC’s on-chip memory controllers mean you don’t need a second CPU to realise the top 2TB capacity - but there is one caveat. RDIMMs support 3,200MHz memory speeds but are only available in sizes up to 64GB whereas LR-DIMMs go up to 128GB sizes but drop speed to 2,666MHz.

Dell PowerEdge R6515 review: Storage permutations

The R6515 offers plenty of storage options with a choice of three backplanes. Base systems have a 4 LFF hot-swap backplane or you can opt for an 8 SFF version if you want to keep the optical drive.

Dell dispensed with the optical drive for our review system, which was supplied with the full 10 SFF bay backplane. A notable feature is it has eight universal bays that support SATA, SAS3 and NVMe devices and two for NVMe SSDs only.

No PCI-E adapter cards are required for NVMe support as the backplane ports are cabled directly to the PCI-E sockets at the front of the motherboard. If you want to mix SATA/SAS3 and NVMe devices, you’ll need to specify a controller card as the motherboard doesn’t have embedded RAID.

RAID choices start with the entry-level PERC H330 Mini card which snaps into the dedicated slot at the rear of the motherboard and supports SATA/SAS3 drives plus stripes, mirrors and RAID5 arrays. Our system sports the top-dog PERC H740P Mini card, which adds RAID6 to the mix and includes 8GB of NVRAM cache plus a battery backup unit.

Dell PowerEdge R6515 review: Power and expansion

Power choices are simplified as the R6515 is only available with dual 550W hot-plug PSUs. Cooling is handled by six fans of the cold-swap variety, arranged behind the drive backplane.

The R6515 can present two PCI-E expansion slots using optional riser cards. These convert the slot nearest the PSU bay to a PCI-E 3.0 version while the one next door employs a larger motherboard connector to provide PCI-E 4.0 services.

The R6515 comes with dual embedded Gigabit ports and you don’t need to lose an expansion slot to expand, as underneath the PCI-E 4.0 riser is a LOM (LAN on motherboard) port. Dell offers a good range of LOM mezzanine cards, and you can choose from dual Gigabit, 10GbE copper or fibre and 25GbE.

Other storage features of note are Dell’s IDSDM (internal dual SD module) which fits into a dedicated slot on the motherboard and provides redundant storage for running a hypervisor. We’ve also included the BOSS (boot optimized storage solution) M.2 SSD card which uses a PCI-E slot and provides mirrored storage using dual M.2 SSDs so you can run an OS without using any front drive bays.

Dell PowerEdge R6515 review: Server management

Remote server management is a class act; the embedded iDRAC9 controller presents a smart web interface offering a vast amount of information on system and component status, power usage and cooling efficiency. Along with full hardware inventory, it provides direct access to BIOS as well as storage configuration, and the Enterprise X5 license activates a remote console and virtual media services.

The new Datacenter X5 license adds telemetry streaming of over 20 hardware metrics for analytics plus thermal and airflow management and tighter access security with automatic certificate enrolment. The only feature you won’t find for any of Dell’s EPYC servers is the System page performance tab, which is still unable to support AMD’s architecture.

The Quick Sync 2 module is a must-have feature for data centre technicians. Using the OpenManage Mobile (OMM) iOS app on our iPad, we connected to it over Bluetooth by scanning the QR code on the pull-out system label and could then view all server information, alerts and health status.

The OpenManage Enterprise (OME) software provides general systems management of all Dell equipment. We run it in the lab as a Hyper-V VM and after discovering the server’s iDRAC9, we were able to manage and monitor it, directly access the controller’s web console and run remote control sessions. We also used the new OME extension for the Windows Admin Center to view health information and hardware inventory.

Dell PowerEdge R6515 review: Verdict

The PowerEdge R6515 is perfect for enterprises that want the highest core density in the smallest rack footprint. It may only be 1U high but the R6515 delivers great expansion potential, a wide range of storage options plus the best remote management tools - and all at a temptingly low price.

Dell PowerEdge R6515 specifications

A former Apple executive dubbed the ‘chief architect’ of its processors has accused the company of reviewing confidential text messages before suing him after he left to start his own firm.

In August, Apple sued Gerard Williams III for alleged breach of contract after he left the company in February to launch his own startup, Nuvia, that’s slated to build processors for data centres. Williams has countered Apple’s claims, however, with allegations that the company monitored his private text messages, according to Bloomberg.

Apple had initially claimed that Williams breached an intellectual property agreement by engaging in business activities that were competitive with or directly linked with Apple’s business activities.

It’s well known, moreover, that Apple may consider designing or customising servers in the future and has dedicated R&D time and effort into optimising server performance internally, the firm’s accusations continued.

Until leaving Apple, Gerard Williams III had led the design of every one of its core mobile CPUs between the A7 chip, found in the iPhone 5S, to the A12X fitted into the 2018 iPad Pro. These chips have forged a reputation for being the fastest CPUs fitted into handheld devices in the market, and possess a unique form of architecture.

Williams, however, has hit back and accused Apple of a “stunning and disquieting invasion of privacy” over monitoring his private text messages. One message, for instance, remarked that Apple would have “no choice but to purchase” his new firm.

"[According to] Apple’s theory, if one Apple employee speaks to (or texts) another employee conveying criticisms of Apple’s strategies or decisions, that discussion is itself a purportedly unlawful ‘solicitation’ to leave Apple,” Williams said in a court filing.

This particular accusation is stark considering Apple has purported itself to be a privacy-centric company in recent years. Sign in with Apple, for example, was launched this year to counter the social media logins provided by companies like Facebook and Google, both accused previously of compromising user privacy.

Apple’s reputation took another hit over the weekend, after it emerged that a feature built into the iPhone 11 tracked users’ locations intermittently in all instances, even when all location tracking features were manually disabled.

Williams has fixed a hearing to attempt to get the case against him dismissed, with his lawyers accusing Apple of desperate attempts to “shut down lawful employment by a former employee”. The hearing is scheduled for 21 January.

IT Pro approached Apple for a statement.

Don't judge Fujitsu's little tower server on size alone. It's certainly small, standing just 405mm tall - but it's a powerful package that's more versatile than many much larger servers.

That's partly thanks to its highly configurable design. The model we tested had a quad-core Xeon E-2134 and 16GB of DDR4 RAM - expandable to a maximum of 64GB - but Fujitsu offers a choice of ten Xeon E-2100 CPUs, with prices starting at just £827 for a base system with 8GB of memory.

There's a wide range of storage choices, too. The entry models support two LFF SATA drives, but for storage-hungry roles, you can specify up to eight hotplug SFF drive bays. If you don't want to go the whole hog right away, you can start with one four-bay cage and add a second extension box when required.

It's worth noting that if you want to use all eight drives, you'll need to add a dedicated RAID controller; happily, Fujitsu offers a great choice of affordable SATA/SAS3 cards. Costing £108, the CP400i adds RAID5 and 50 to the mix, while for around £190 the top-dog EP420i supports RAID6 and 60 arrays, as well as an optional battery backup pack to protect its 2GB of onboard cache.

For faster access, the motherboard also has dual M.2 SSD slots, one of which supports both SATA and NVMe modules. Invest in one of Fujitsu's PRAID EP540i or EP580i dual-function controllers and the server can support up to four high-performance NVMe SSDs.

Expansion in other areas is good, too. Four PCIe slots and a pair of Gigabit Ethernet ports are partnered with a dedicated management port for Fujitsu's iRMC S5 management controller. This can be used in concert with Fujitsu's free ServerView Suite software, but that's primarily aimed at large businesses managing multiple server. Smaller companies will find the System Monitor utility more practical: its browser-based console provides at-a-glance details on critical components and power consumption, with an automated system update service. Upgrading the Standard licence to an Advanced one enables OS remote control and virtual media services as well.

To power it all, starter systems use a fixed 250W PSU, but our configuration included a 450W hotplug PSU with room for another alongside. The second bay also accepts Fujitsu's battery backup module (FJBU) which costs around £300 and includes a Windows management tool that gracefully powers the server down in the event of an outage lasting longer than four minutes.

While the TX1320 M4 may be modest in size, it's no lightweight: the steel chassis is reassuringly solid and offers good physical security, as the side panel can be padlocked shut and the two sections of the front bezel locked with a single key.

Open it up, and everything is nicely accessible. LFF drives are fixed in their cage with thumbscrews and are easy to remove, while the slimline optical drive above is connected to a dedicated SATA port on the motherboard. A spare bay next door can take an optional RDX backup cartridge drive, and the motherboard has an embedded USB port located right next to it.

Thanks to a tidy internal design, not much in the way of cooling is required. There's no need for a rear-mounted fan; instead, the CPU has a smart active heatsink with a fan neatly sandwiched inside it. The LFF model also has a single cooling fan servicing its drive bays, while the SFF model gets a clip-on fan module for each four-bay cage.

This sparing use of fans keeps the server admirably quiet. We measured sound levels of only 37.6dB from one metre in front, so you don't have to worry about nearby workers being disturbed by the noise: it's fine for a small office, or even a library.

In all, the Fujitsu Primergy TX1320 M4 is ideal for space-constrained businesses. Compact yet capable, it's great value, and can be configured to suit a wide range of demanding workloads.

Cloud computing has revolutionized business operations across the globe but in these uncertain times, there is a clear need for maintaining on-premises IT services. Disasters, wildfires, climate change, geopolitics, cyberattacks, systems upgrades, software updates – they've all caused significant cloud service outages, and their increasing frequency means it's not if, but when, they'll next go offline and leave your business in the lurch.

There's a lot to be said in favor of on-premises servers. With the right supporting infrastructure in place, they can deliver 24/7 data availability and put you in full control of your services and business apps. 

Cloud providers generally adhere to the 'shared responsibility model' where they look after the security of the infrastructure, data centers, and server hardware leaving customers to handle cloud applications, data, operating system, and access security. Moving services to on-premises servers means you'll be in control of all aspects of their security, including data governance.

You'll also have total equipment ownership and will be free to make any customisations you want, whenever you want and store your data in more secure and cost-effective ways. True, your capital expenditure (CapEx) will be a lot higher, but compared to ongoing cloud service subscriptions, your operating expenses (OpEx) over the long-term could be significantly lower.

1. Plan for future server capacity

The diverse range of server solutions on the market makes it easy to get the right systems for your current needs but you should plan for the future so they have plenty of power on tap to cope with the inevitable increased demand. Servers are long-term investments and choosing the right ones will ensure they can stay with you for years to come.

Internal server storage can get eaten up quickly so it's essential you plan capacity for the future to avoid costly upgrades and unnecessary downtime. File sharing, private clouds, and database services will grow over time so consider servers that have spare bays for adding more storage devices – it's a lot quicker and cheaper to expand an existing RAID array with extra disks than it is to replace the entire server.

CapEx comes into play here again as once you've purchased your on-premises storage, its OpEx will be negligible. Cloud storage, on the other hand, incurs continual expenditure with many businesses forced to purchase more capacity to keep up with demand finding this is costing them a lot more than they'd bargained for.

2. Decide which servers to buy

Server buying decisions will be based on many factors including the type of workloads, performance requirements, expansion potential, latency, and even their physical location. Space-constrained small businesses looking for their first server or a timely upgrade will find rack models an appealing choice. 

Low-profile 1U rack servers are ideal where businesses want to separate out their workloads onto rack cabinets full of dedicated systems while 2U models provide more expansion potential and higher storage densities. Offices are expensive to rent and the latest rack servers pack a remarkable amount of features, processing power, and storage into the smallest of spaces. 

They're versatile too, as they can be installed in a small footprint rack cabinet with some small enough to slip into a wall-mount enclosure. We've reviewed most of Dell's latest Gen16 PowerEdge rack servers to help you make the right choice.

If you're looking at pushing IT services out to remote offices and other locations where data will be processed on-site then tower servers may be the best choice as they don't need any additional infrastructure such as rack cabinets. They're also ideal for rapidly expanding small businesses looking to replace their legacy PC-based network with a purpose-built server.

Edge servers are also worth considering as they deliver a powerful hardware package in a space-saving chassis that's designed to cope with harsh environments such as factories and shop floors. We've reviewed a range of edge servers including Dell's PowerEdge XR11 and Lenovo's ThinkEdge SR450 models.

3. Configure server networking

The supporting network infrastructure is just as important as your choice of server as this must provide a solid foundation for it to deliver its applications and services. Ethernet cables are the networks' arteries and it pays to use the best quality to ensure connection speeds are unaffected and future demand can be met. 

If you're planning a new structured cabling project then the most cost-effective choice is Cat6a cables as these support 10-Gigabit (10GbE) speeds up to 100 meters – you can ignore Cat7 and Cat7a cabling as they were never standardized. If you can afford it go for the latest Cat8 cabling as this has four times the bandwidth of Cat6a, uses the same RJ-45 connectors, and supports 25GbE and 40GbE speeds over 30 meters and 10GbE over 100 meters making it ideal for switch-to-switch and server to top-of-rack switch connections.

Network switches are just as important so your budget should extend to sourcing the fastest you can afford and don't forget their port count either. As your network expands, ports will get used up quickly so if you think a 24-port switch will do the job now then get a 48-port model instead so you have plenty of headroom.

Many businesses have mobile workforces so good wireless networks are essential to allow them to access on-premises and cloud services. There is a vast range of fast wireless access points to choose from, but a feature across all is a requirement for Power over Ethernet (PoE) so you should consider adding PoE switches to power them.

4. Consider going hybrid

Going on-prem doesn't mean you should ignore the cloud. Far from it, as many businesses are seeing the benefits of a hybrid approach where the public cloud is considered an essential component of their IT strategy while more demanding apps are maintained in a private cloud.

There are many reasons for these decisions although increasing public cloud ingress, egress, API call, and data transfer charges can make it more cost-effective to run certain workloads in-house on dedicated servers. 

In fact, an increasing number of businesses that initially took a cloud-only approach are repatriating data from the cloud in varying amounts as they see the cost benefits of a hybrid infrastructure.

Combining cloud services and platforms with your own in-house servers allows you to keep mission-critical services at your fingertips but push less essential services and applications into the cloud. The cost benefits alone are a strong argument in favor of this approach and make a sound business case for running essential applications and services on your own on-premises servers.