A Broadband Internet Technical Advisory Group Technical Working Group Report.
Please direct comments on the substance of the report to [email protected].
Executive Summary
It is time to update our understanding of the primary factors directly affecting end-user Internet performance. What we have learned is that high throughput alone is not sufficient. Latency is also a critical factor in providing a high performance Internet connection. But that is not latency as we have traditionally understood it; rather, it is what we now define as ‘working latency’.
For over thirty years, industry has collectively missed a key factor that drives end users’ Internet quality of experience (QoE)! The industry has historically been focused on increasing bandwidth, which has been straightforward, easy to understand, and indisputably made dramatic improvements to end-user QoE.
But we now recognize that it is not just greater throughput that matters, but also consistently low latency. Unfortunately, the way that we’ve historically understood and characterized latency was flawed, and our latency measurements and metrics were not aligned with end-user QoE.
The industry is now coming to a new understanding of what latency really is, the role that latency plays in QoE, how latency occurs in any access network technology, and how to improve it. It is critical that policymakers understand this significant shift in our understanding of the Internet, of coming improvements and new capabilities, and what this will mean for end users. We envision not only improved QoE for all of the applications that are used today, but a foundation for entirely new classes of applications to be created.
Policymakers should read this report to develop a new understanding of latency, and industry should implement and deploy technologies that reduce latency.
A key observation of this report is that one of the most impactful (and solvable) sources of latency and latency variation affecting Internet users is buffering delay. Buffering delay arises from an interplay between application behaviors and network equipment implementations, and has been mistakenly (although sometimes intentionally) ignored in many latency measurement campaigns, which has served to prevent it from being identified as a root cause of application degradations. Measurement techniques that seek to include the effects of buffering delay in the measurment are sometimes referred to as latency under load or working latency tests.
This report will explain and explore latency, including idle latency, working latency (latency under load), components of latency, how some of those components vary by access technology and/or protocol. It investigates the effect that latency can have on the user QoE for certain types of applications such as video conferencing and remote learning, gaming, and more. Additionally, this report will examine measurement methods and metrics for characterizing latency, including recommending a shift in the primary way network latency is measured and reported. Finally, this report will look to the current state of the art and the future to explain the importance of Active Queue Management (AQM) and other low-latency networking innovations in reducing latency, enhancing broadband performance, and fostering new types of applications in the future.