An Experimental Evaluation of 360-Degree ABR Video Streaming over mmWave Wireless Links

Video streaming is a key technology for the envisioned Metaverse. Present wireless technologies have limited throughput, which limits the quality of video content (e.g., 360° video) that can be delivered to mobile clients. To address this challenge, wireless carriers have begun to deploy millimeter wave (mmWave) networks, which have high throughput potentially in the multi-gigabit per second range. However, mmWave wireless is highly sensitive to blockage and mobility that induce dramatic variations in link-level throughput. This can in turn penalize the video quality of experience delivered to clients. This adverse effect can potentially be mitigated by the use of adaptive bitrate (ABR) techniques that dynamically adjust the bitrate of the video content based on network conditions and device capabilities. However, existing studies of ABR video streaming performance over mmWave wireless links assume channels with low throughput, use simulations (with synthetic or measured traces) for performance evaluation, and/or conduct large scale experiments that mask client performance at the individual level. In this paper, we develop a testbed to experimentally assess 360° video ABR performance of three prominent methods over high capacity (high throughput) mmWave wireless links for a variety of channel conditions. We show that the fundamental tradeoff between stalls and visual quality becomes much sharper at higher throughput, and that existing ABR methods do not achieve substantial gains in one without losses in the other. For instance, we find that complex ABR strategies which outperform simpler strategies in terms of visual QoE also increase stall time from less than 0.5 seconds to nearly 4.