You've just had a brand-new broadband connection installed and are ready for anything the internet can throw at you.
The network carrier's salespeople promised you blistering bandwidth, but your service seems slower than it should. Most of us have been in this situation at some point.
Rather than relying on your carrier's marketing figures, it pays to measure your network performance yourself. We’ll explain how here.
Bandwidth versus throughput
When we talk about performance, what exactly are we measuring? It's important to understand the difference between bandwidth and throughput.
Bandwidth is the connection's capacity for carrying data. Think of a high-bandwidth connection as a six-lane highway compared to a two-lane road. The former can hold more traffic than the latter.
Throughput measures the speed at which traffic is passing over the network, and is generally reported in data units per second — typically megabytes per second (Mbps).
High bandwidth often means high throughput. However, just as a six-lane highway suffers jams that bring traffic to a standstill, factors can decrease throughput on a high-bandwidth network. These include contention between many clients using that connection, which can slow individual users’ throughout to a trickle.
Another potential issue is high latency. This is the time it takes for a packet to reach its destination, and it's critical to some applications like voice over internet protocol (VoIP). Having lots of devices on the connection route or other bottlenecks along the way can increase latency, affecting network speed.
If latency varies too much across a session, you might experience jitter, which is a problem for sensitive applications such as VoIP or video conferencing. Jitter stems from factors like network interference or poor network signals. You might find this on Wi-Fi or cellular connections, but electrical activity from storms and even high-powered electrical motors can affect copper cabling.
Measuring broadband throughput
When people talk about measuring bandwidth, they're often really talking about throughput. Most throughput tests use a similar method, sending a series of packets to a destination and measuring the response time over a set period to get an average.
Broadband throughput measurement tools often give you different metrics. Some wide-area broadband connections will often offer asymmetric bandwidth, meaning you get more capacity for data downloads than for uploads. The best testing tools will break these out for you.
Another common metric is the ping speed, which measures the round trip for packet delivery. As a rule, you want ping speeds of 20ms or less for VoIP calls, although you can get tolerable results at up to 150ms.
You'll find plenty of broadband testing tools online, some of which come directly from carriers like Comcast and Xfinity. Still, you can also get independent versions like Ookla's Speedtest.net (which also happens to power AT&T's network testing site).
Broadband speed testing tools won't all deliver equal results, which is why it's best to use either your ISP's testing tool or an independent speed testing tool like Ookla's.
For example, the Comcast Business testing tool put my latency at 67ms, whereas Speedtest said it was at just 4ms. Similarly, Comcast said I uploaded at 22Mbps, but Speedtest put it at over 10 times that. Comcast measured my speed by connecting me to a host over 2,400 miles away, presumably on its own network, while Speedtest connected me to a server in a city 62 miles away and on my ISP's network. These things matter when it comes to network throughput.
The configuration on your side of the router also matters when assessing throughput. Even though your broadband connection might be consistent and predictable from your router outward, the devices on your network and how they're configured will affect the last few yards between your router and the endpoint device you're testing on.
Testing your broadband throughput using your regular network setup will give you a picture of your throughout under real network conditions. Then you can test it under ideal conditions by removing as many variables as possible from the testing process.
One way to do this is connecting directly to your internet modem or gateway via an Ethernet cable, circumventing any switches, hubs, or Wi-Fi. Use a machine with a gigabit Ethernet network interface card (NIC) to ensure the machine can support higher-bandwidth connections like fiber.
LAN tests
Now is also the time to test the bandwidth between various points inside the LAN, especially if you notice a big difference in broadband throughput when testing under real network conditions with lots of other clients using the network.
Reading the NICs' on your sending and receiving machines and the supported speed of any network equipment in the middle will give you a theoretical bandwidth.
In reality, though, your mileage may vary depending on factors, including how many other devices are on the network, how much bandwidth those other devices are consuming, and what network equipment, such as hubs or Wi-Fi connections, are supporting them. All it takes is an overloaded switch uplink port to lower your real network throughput.
LAN throughput testing tools on a LAN include NetStress, a free GUI-based tool that runs on both endpoints in a connection, enabling you to test the bandwidth between them.
NetStress only works with Windows, whereas iperf, a free command-line tool that also monitors throughput, works on various systems, including desktop Linux, OSX, Android, and even BSD variants. If you want iOS capability, consider TamoSoft Throughput Test, which works on modern versions of Apple's mobile OS alongside Android and Windows.
Measuring Wi-Fi throughput
Wi-Fi connections are among the most variable when it comes to bandwidth and speed. While Wi-Fi standards offer theoretical bandwidth figures, the physical office environment can affect real-world conditions.
Wi-Fi access point placement, antennae direction, walls in the building, and the distance of the endpoint from the access point all impact Wi-Fi performance.
Even the endpoints themselves can have a bearing on test results. Testing using a cellphone app could yield varying results depending on the cellphone's internal Wi-Fi transceiver capabilities, which can vary between models. Many admins will value a mobile app’s convenience regardless, in which case there are plenty of Wi-Fi network analyzer apps available for Android and iOS.
One way to test Wi-Fi network bandwidth under ideal network conditions is to connect two endpoints to a Wi-Fi access point with as few other connections using it as possible. Ideally, one of these devices would connect directly via Ethernet to a wired port on the Wi-Fi access point if it has one. This gives at least one of the endpoints in the conversation a consistent connection. Then, you can run a LAN-based bandwidth measurement tool between the two devices.
You could then repeat the exercise with the wireless endpoint at various distances from the access point to assess the throughput change.
How much bandwidth do you need?
You can use these techniques to test the throughput, latency, and jitter between different points in your network, including between your endpoints and your internet gateway, giving you a clear picture of any bottlenecks to address.
Armed with this information, you can take care of chokepoints in your LAN infrastructure and ensure you're using your bandwidth as efficiently as possible. Then look at your IT roadmap to determine whether you have enough capacity to support any planned changes.
Assess each application according to its projected throughput requirements per user and multiply this by the number of expected concurrent users.
What should you do if you don't have enough throughput to support your plans? Buying more wide-area bandwidth is an obvious step. There are other potential options too, including WAN optimization, which uses deduplication, compression, and caching techniques to squeeze more out of the available pipe.
Other options include quality of service (QoS) technology to prioritize latency-sensitive traffic and more critical applications such as VoIP and video conferencing.
If you need more WAN bandwidth, it's worth examining your network usage profile before negotiating your contract. If traffic only peaks above acceptable thresholds at certain times, consider bandwidth-on-demand contracts that let customers burst traffic during peak periods.
Finally, it pays to test your network throughput more than once. Try sampling the test several times over 24 hours to generate an average. This will also alert you to any large variations that could indicate a problem between you and your provider.
Performing these tests every few months will allow you to trend your network throughput over time, alerting you to any emerging networking issues before they become a problem.
