Most users only consider one measure of performance when it comes to their Internet connections and that’s speed. It’s easily relatable and understandable — if we have a 10 Mbps connection, it takes one second to download a 10 Mb file. Other metrics, such as bandwidth and latency, aren’t as well understood and, apparently, that’s a problem. Researchers from the University of Illinois at Champaign-Urbana and Duke University say that solving network latency problems could allow networking at the speed of light.
Theoretically, we should already be able to transmit data around networks at nearly the speed of light, but in practice, that’s not the case. The researchers found network latency is often two orders of magnitude—100 times—higher than what it should be. They propose a change to networking infrastructure that could remove one of those orders of magnitude entirely, reducing latency to a tenth of current times.
Latency is actually pretty simple to understand: it’s the measure in milliseconds of how long it takes a packet of information to get from one place on a network to another. For example, we all understand the concept of the speed of sound — if someone fires a gun far away, we see the puff of smoke before we hear the bang. That’s latency. As it relates to the Internet, we’re most familiar with latency in terms of hesitation in online gaming, or wireless speakers out of sync with streaming video.
In the example of sound, latency is caused by how fast sound waves can travel through the medium of air. If you have to transmit a noise and wanted to reduce the latency, you can speed up transmission by altering the medium the sound moves through — sound moves much faster through water than air.
There’s really no difference between transferring sound and transferring network data: change the transfer medium, and you change the latency. Just as sound can travel through water faster than air, data can travel faster through microwaves than traditional networking media, even fiber. Microwaves are already used to transmit data between financial markets and exchanges because speed is critical, and small delays can have huge consequences. This technology, the researchers say, is the solution we need.
Wait a Minute…
Why not just replace fiber with microwave networks? The authors say there might not be enough bandwidth in the microwave spectrum to make such a thing possible. Microwaves trade excellent speed for comparatively poor bandwidth performance. The researchers’ solution is simple: transfer latency-sensitive data via microwave network and all other data via fiber or other networks. This would establish twin Internets, in effect, with the new, microwave system known as “cISP”—c for the speed of light constant.
The authors estimate that building cISP would cost $253 million and cost $96 million in yearly operational costs. That sounds like a lot of money, but when viewed as a national resource, it’s really not. The FCC is spending more than six times this amount to increase broadband availability in rural areas. San Francisco estimated that it would cost more than double that figure just to bring broadband to every resident of the city.
Although microwave technology has bandwidth limitations, it still adds bandwidth overall, and the authors propose a cISP bandwidth target of 80 Gbps. That would cover the most time-sensitive content, like video conferencing, currency trading, and so on.
Latency doesn’t really matter for some online content — consider email, Internet of Things (IoT) sensor technology, and that old standby, looking at cat pictures. In these examples, a slightly larger delay is more acceptable. For example, it doesn’t matter if it takes online video a few more seconds to reach your house so long as your connection speed allows for good playback once it does begin. For this content, the network infrastructure we’re used to should be sufficient.
How’s Your Plan Compare?
If you’ve noticed slower gaming, video chat, or other latency-sensitive applications, you can run an online latency test to see how your current plan measures up. If you’re not satisfied, maybe it’s time you looked for a plan that’s better, cheaper—or both.
Photo Credit: Thomas Hawk/Flikr