The Waze application has revolutionized our driving experience. Instead of being backed up at red lights and getting clogged in traffic jams, Waze offers a platform to get to our destination using the fastest and most effective route.
We face a similar challenge in our Internet experience and there have been many efforts to similarly use rerouting and load distribution schemes online. So far, they have largely been insufficient. But now there is finally hope with a mechanism able to avoid these “traffic jams” and have a real impact on our Internet journey.
L4S looks to limit the buffering and delays we all experience when surfing the web, having video chats and playing interactive, on-line games. It could be the key that unlocks the full potential of future connectivity.
L4S stands for “Low Latency, Low Loss, Scalable throughput,” and it drastically reduces and eliminates latency experienced by packets traveling across the Internet. As such, it could revolutionize extended reality, video conferencing, multiplayer gaming and numerous other applications that require low-latency connections.
The result of initial pioneering Nokia Bell Labs research that later included a larger community of enthusiastic supporters and contributors, L4S looks to redefine and optimize collaboration between the network and its applications in a way that reduces existing latency levels by factors of between 10 to 1000.
It was recently granted RFC status by the Internet Engineering Task Force, setting it up to become an industry standard.
For the man behind the breakthrough technology, it’s the culmination of the decade-long journey that began on a different project and evolved into a multi-disciplinary mission of looking into how to solve internet congestion in an age of exponential growth in capacity demand.
“Latency is the blocking factor,” said Koen De Schepper, a veteran Bell Labs research principal with more than 30 patents to his name. “Over the past 30-40 years, the Internet has organically grown with the best possible unilateral improvements. There have been collaborative improvements in the past, but the benefits couldn’t outweigh the costs of rolling it out. Now we have found a way.”
An early curiosity in Belgium
It’s that curiosity that got him interested in science in the first place. As a kid growing in Antwerp, Belgium, long before the Internet existed, he said he always wanted to find out how things worked. That mystery led to early technical courses in electricity and later advanced degrees in electronics and software engineering.
De Schepper has lived his entire life in Belgium and spent his entire professional career at the same company owned by Alcatel and later Alcatel-Lucent before they were incorporated into Nokia. But his areas of work have varied wildly, from building code generators to large development platforms, from library development to architecture and research. Throughout, he has always kept one goal in mind: providing real solutions to application developers.
Today, he leads the L4S research at Bell Labs in the Network Systems and Security Research lab. His research focuses on providing network and application functions to improve scalable and low latency content delivery.
His particular interest in transport protocol and cross-layer network interactions led to L4S. But he said it was his early days in tech that truly inspired the breakthrough. That was back when he worked in platform development, and he had direct contact with large teams of application developers and a good sense of their needs. As a researcher, he said he has maintained the same outlook of always thinking about what will have the best outcome for platform users.
“It’s important for me to have freedom to improve things and build things that make life easier for other people,” he explained. “You can always enforce rules in platforms and ‘make’ people use them. Or you can make things easy and beneficial to use so that people ‘want’ to use it.”
The evolution of the internet
When it comes to the internet, there are two key factors that make up the user experience: throughput and latency. The focus has traditionally been on the fastest throughput, increasing download speed and capacity as much as possible.
Download speed can only be improved with higher throughputs that reduce waiting times from minutes to seconds. That process has largely been successful. But when speeds or waiting times need to drop well below a second, latency becomes more important.
L4S flips the dynamic by focusing on the lowest latency, which is key for next generation use cases like extended reality and interactive video conferencing, and mission-critical services such as remote driving and medical procedures. It’s one thing to have a one-second delay watching a video and quite another when remotely controlling a critical machine or vehicle.
The secret to significantly improving end-to-end latency lies in tackling the single biggest and often most overlooked source of latency and jitter: queuing delay. That’s the stage where packets wait in buffers before being forwarded.
End-to-end latency includes three factors: propagation, interface and queuing delay.
Queuing delay happens when packets wait idly in routers and modems, for example. As users and bandwidth-intensive applications send more and more traffic through the network, these queued data packets cause network links to get “clogged.” Packets in that congested pipe take longer to reach their destinations, and some packets even get lost in route when the buffer overflows.
Today, such delays cause videos to stall, people to start talking over one another in videoconferencing or frames to be skipped in cloud-gaming applications. In the future, though, the consequences could be far more devastating, with even the slightest delay causing a remote-controlled drone to crash.
L4S addresses that, removing the need for big buffers.
De Schepper said that was just touching the surface of all the potential use cases.
“I’m confident there is enough creativity for applications we can’t even imagine yet,” he added.
A paradigm inversion
Traditionally, network engineers have facilitated waiting queues because applications have needed long waiting lines to be processed with low packet loss for improved user experience.
L4S, on the other hand, removes all the reasons why the network needed to build these long waiting queues, resulting in an optimized user experience and flipping the previous rationale.
“So, from a network design perspective, L4S is not a paradigm shift, it is a paradigm inversion,” said De Schepper.
To go back to the street traffic analogy, De Schepper said that L4S, like Waze, doesn’t tell you what route to take for an optimal journey but rather when to leave and how much load to carry.
“In the existing internet, the current rules for knowing when and how much to send are based on how long you must wait in line and how much of your load gets lost. Everyone will run into traffic lights and must wait in line,” he said. “L4S can tell packets when to go to avoid any red lights along the way.”
The result, he said, can easily be a factor of one thousand latency improvement.
“We aren’t trying to reroute traffic, we’re mostly telling them when to go to avoid a queue,” he continued.
De Schepper, a father of three, and grandfather of one, has many creative hobbies, including winemaking, painting, robotics and various home improvement projects. In fact, he hopes to see the day when L4S becomes a mainstream commodity that he can enjoy in the comfort of his own home.
“Of course, I’m eager to use it myself,” he said. “I hope to have a video wall that can connect homes together. An internet window where people could be side-by-side as if they were really sitting next to each other. L4S has the capability to make that feel like a face-to-face meeting.”