An examination of live video streaming performance when playing with different live delays.

What is it about?

Our paper examines how live video streaming performs when playing with different delays from live, across a range of different network conditions. The study focuses on three different adaptation methods used by the widely used dash.js web-based video player. We conducted experiments using a testing environment to mimic real-world network situations. The goal was to understand how different delay, or latency, targets affected the viewer's quality of experience and how well each adaptation method met those targets. We found that the default method in dash.js provided the best overall experience. Although the other methods could deliver higher video quality at faster speeds, they often caused interruptions in the playback. The study also investigated a specific adaptation method called L2A-LL, which initially performed poorly in our tests. We made modifications to significantly improve its performance. Additionally, we discovered that certain player settings could make the performance worse for low latency.

Featured Image

Photo by Guillaume Jaillet on Unsplash

Photo by Guillaume Jaillet on Unsplash

Why is it important?

It will help with a better understanding of how best to deploy low latency video streaming. We found that dash.js generally succeeds in maintaining its target latency. Under favourable network conditions, the Dynamic and modified L2A-LL algorithms maintained a reasonable quality of experience as latency was reduced from 15s to the 6-8s region. We observed that all algorithms struggle under more challenging network conditions. We saw that significant differences in performance were apparent between different dash.js versions. The FastSwitching feature was found to not benefit low latency QoE but generates additional traffic

Perspectives