Anatomy of Multi-gig Speed Testing
Akin to the struggles of medieval alchemists seeking the mythic philosopher’s stone, accurately testing your Internet connection’s speed requires perseverance. Testing 1Gb connections has been challenging enough but with multi-gig residential fiber deployments becoming more common we’re seeing consumers express frustration at their inability to prove that they’re getting what they’re paying for.
The established tool for testing home Internet connections has been speed tests in your browser on your home computer but many things can impede that solution. For example, many people no longer even have a wired home computer and speed tests over WiFi have limitations making even 1Gb connection tests impossible to achieve. Speed tests were designed years ago to test at much slower speeds and as capacity has increased new challenges have arisen.
The difficulties of multi-gig speed testing
Let’s start by summarizing how traditional speed testing works. Within your browser a file is uploaded and downloaded then your Internet connection speed is determined by measuring how quickly that file is moved. Simple enough but in reality a speed test shows the highest speed you’re capable of depending on your worst bottleneck, which often isn’t your Internet connection once you have a multi-gig port. After all, we can’t blame the road when we arrive late to work due to car problems. In such a case your home router or computer is like a car.
Ultimately, many bottlenecks can exist when you’re in the upper limits of Internet speed:
- Your router (most consumer grade equipment can’t reach multi-gig speeds)
- WiFi limitations (even the newest WiFi can’t reach multi-gig speeds in real-world use)
- Viruses/spyware/adware slowing down your computer
- A slow, older computer
- Limitations of your browser(s) (they weren’t designed to move huge files quickly)
- Other applications running on your computer
- Other devices connected to your home network using bandwidth (streaming shows or playing games)
- The quality of your Ethernet cables
- Packet overhead from TCP
- Port size or processor power at the speed test server
Therefore, we’ve known for a while now that the best option for most people wanting a reliable measurement of their Internet speed is to bypass most of these potential bottlenecks and either test from a multi-gig capable WiFi router using software designed for the purpose or use an app on your workstation that’s wired directly to your fiber equipment. Interestingly though, even when we followed these established criteria we weren’t always seeing expected speeds in the 7–9Gb range. The rest of this article recounts our experiments to figure out the most reliable and accurate solutions for customers to test their high speed connections.
Networking fundamentals
Let’s dig a little deeper into the technical reasons why accurate multi-gig testing is so difficult to achieve. Firstly, Internet protocols were not written for the high speed network XMission has deployed. Back in the 1970s and 1980s, a 1Mb connection was massive and the sort of traffic early DARPA engineers needed to transmit was significantly less than that now deployed to the average American home. They developed TCP/IP (Transmission Control Protocol/Internet Protocol) to suit their needs at the time which primarily required accurately sending data packets from point A to point B. For example, if an expected packet doesn’t reach point B in a timely manner a request will be sent asking for it again. If you need to ensure file transfer integrity such features are essential whereas speed is treated as a lower priority.
A key bottleneck we’re seeing with multi-gig testing stems from this concern for file integrity. Most speed tests use TCP but when every packet needs to be confirmed as received at point B latency emerges as an unexpected bottleneck. For example, on a 10Gb connection using quality gear we might only see 3-5Gb from a speed test server to a computer browser because browsers weren’t designed to transfer files so quickly.
As we dug deeper, we also found download speed issues can be impinged due to latency (e.g., ping). That is, distance and the number of routers between you and the speed test server cause higher latency, especially if you’re using TCP with the test. Therefore, you always want to find the closest server available, which is why XMission hosts one internally at speedtest.xmission.com and we recommend customers use it for the best results.
Until recently, XMission ran our speed test server with a 10Gb port. Since the vast majority of our fiber customers had 1Gb connections we considered it sufficient and educated our relatively few multi-gig customers to manage their expectations. We expected that 10Gb customers could still get close to the full available speed but after some complaints we started to wonder if upgrading the hardware and port speed would have an impact. As we neared the end of 2024, we noticed we were acquiring far more multi-gig customers (esp. 2.5 Gb) and decided to build a “hot rod” speed test server with a 100Gb port and lots of CPU power and RAM.
We expected to see improved multi-gig speed measurements. After initially deploying the new hardware and tweaking some settings we started to run tests to the beefy new server but were quite disappointed: no noticeable change testing multi-gig speeds. We’ll provide some of those results in the XMission Lab post but we didn’t manage to reach 9Gb speeds until we upgraded an office switch to support 10Gb ports and bought quality 10Gb USB network adapters to run on powerful laptops. Even then, we only saw such results when running advanced speed testing solutions like iPerf, and the Ookla desktop app. These are best case “lab” scenarios where the speed test server is in a nearby room and all other potential bottlenecks are bypassed. It should also be noted that regardless of how you test and what speed test server you use, the result is only accurately showing you the speed between you and the speed test server, nowhere else.
Customer premise equipment
A big struggle for multi-gig customers has always been the lack of decent, affordable solutions. For this reason, XMission uses the Calix hardware and app for our XWi solution, which we chose to provide after seeing customers struggle with most consumer grade routers. While we had finally confirmed our new speed test server was capable of providing 10Gb speeds (minus the cost of packet overhead), the excellent 10Gb capable Calix WiFi routers we had started leasing to 10Gb customers were still only internally testing at about 3Gb using both available testing options: TCP and iPerf. We knew these quality devices could handle nearly 10Gb of traffic across multiple streams but it was challenging for our techs to explain this to customers paying a premium for the higher speeds who couldn’t easily see verification via speed tests, even from the quality router itself. Calix automatically runs speed tests from these routers daily and graphs the last seven days’ worth so we can easily reference and verify performance across a week’s worth of days on multiple devices. This 3Gb limit was shown across all Calix 10Gb devices, which left us in a challenging position since we had already determined that the only other reliable way for customers to get an accurate speed test was to directly connect a wired computer booted into Windows safe mode with networking to the fiber ONT and run iPerf or the Ookla app.
Solutions
Fortunately, our vendor Calix recently added a third option, using UDP, the same protocol used for VoIP, games, and other latency-sensitive solutions that can afford to lose a packet sometimes so now we’re seeing results from the router in the 7Gb range. We suspect the new UDP testing option shows better results because the included processor in these routers bogs down due to the issues we’ve already mentioned with TCP. All new 10Gb customers can get the Calix router included at no additional charge so finally we have a simple solution for them to verify they truly are getting the speeds they expect. While we ended up realizing that the speed test server upgrade was unnecessary, over time as more and more customers upgrade to multi-gig speeds it’s nice to have plenty of head room for accurate testing.
XMission now offers multi-gigabit connections on Southern Ute Indian Reservation XMission Labs: Speed Test Lab