L4S Technology: A New Congestion-Control Solution for Latency 

Low Latency Blog

Greg White
Distinguished Technologist

Aug 15, 2023

In a digital world where every second counts, technologies that enable smooth, efficient transmission of data are paramount to ensuring the quality of our online experiences. Reliable connectivity is a must, and the need for it grows exponentially more essential every day, particularly as more applications harness the power of 10G.

The 10G platform is a game-changing, multigigabit network made possible by DOCSIS technologies. It will deliver faster speeds, enhanced reliability, better security and lower latency. Low latency is especially critical for real-time communication applications because it helps create user experiences that are free of delay and disruption.

Many of the applications that we use every day weren’t designed with low latency in mind, and they actually cause network congestion that can disrupt our real-time communications. That’s why we need solutions that can relieve networks of this congestion by combating latency, jitter and packet loss. Low-Latency, Low-Loss Scalable Throughput (L4S) technology is a part of that solution.

What Causes Network Congestion and Latency?

It’s important to understand the source of network congestion and the reason for it. Although it’s well known that latency increases when congestion increases, what’s less commonly known is why it happens. Contrary to what many people think, it’s not the result of too much traffic and too little available bandwidth.

In reality, most of the applications we use every day use congestion control. What that means is that applications are constantly adjusting their sending rates, aiming to send as fast as they can and backing off only when they detect congestion. But congestion-control mechanisms that applications use haven’t evolved significantly since the early days of the internet in the mid-1980s. Those algorithms, such as TCP Reno and CUBIC, rely on the network to provide deep packet buffers and then drop packets when the buffers overflow. The algorithm ramps up, causes delay and packet loss, backs off and ramps up again. As a result, those mechanisms can introduce latency, jitter (also known as latency variation) and packet loss — not only to themselves but also to other applications using the network at the same time.

What Is L4S?

L4S is a core component of the Low Latency DOCSIS specifications, which many other networking technologies are also adopting. It enables a new congestion-control mechanism for capacity-seeking applications wanting to optimize their sending rate, while dramatically reducing network latency, jitter and packet loss.

L4S is ideal for applications that are optimized by high data rates, consistent ultra-low latency and near-zero packet loss — including cloud gaming and virtual reality/augmented reality (VR/AR) applications and high-quality video conferencing. L4S is also beneficial for other applications that are latency-bound, such as general HTTP traffic.

How Does L4S Work?

The end-to-end L4S solution provides high-fidelity congestion feedback from the network bottlenecks to the applications being used. The process involves applications implementing a new congestion-control algorithm that can understand that feedback, adjust their sending rates with better precision and fully utilize link capacity without causing latency and packet loss.

The L4S mechanism can be used by any application, such as TCP or QUIC, as well as real-time applications that use UDP or RTP.

A standardized solution defined by the Internet Engineering Task Force, L4S has already been adopted by cable broadband networks and is supported in the DOCSIS cable modem protocol. Work is underway to implement it in 5G and Wi-Fi networks.

I delve much deeper into L4S in the video below. Watch to learn more about how it's being implemented, the results of recent testing and performance findings, and more.