A Deep Dive Into Live 360-Degree Video Latency

Latency is a critical metric for assessing user experience in 360-degree video streaming. Encompassing both viewport switching latency and end-to-end frame delay, it determines the system's responsiveness and fluidity, directly affecting the user's perception of immersive continuity and content timeliness. In addition to the high resolution, large data volume, and complex processing inherent to 360-degree video, latency is further influenced by factors such as video content dynamics, network conditions, and client device capabilities. A clear understanding of how these factors affect performance is essential for reducing latency and enhancing the user experience. Although existing studies have decomposed these latencies, they fall short of providing a comprehensive and in-depth analysis of how above three factors affect individual components of latency. In this paper, we build a complete end-to-end streaming system and conduct frame-level measurements to analyze the impact of aforementioned factors on both viewport switching latency and end-to-end frame delay. We find that server response emerges as a major bottleneck for viewport switching and can be further exacerbated by dynamic video content. Moreover, limited client-side rendering capability is another significant source of delay. Additionally, server-side transcoding and camera latency are identified as primary bottlenecks in end-to-end frame latency. Although network latency is significantly lower than processing delays, the cascading effects triggered by elevated delays should not be overlooked. While shortening the keyframe interval or lowering video quality are effective in reducing latency, we further propose additional directions to inform the design and optimization of low-latency 360-degree video streaming systems.