Server-Side Segment Selection for Low-Latency Streaming—S4S
December 16, 2021
Guillaume Bichot and Nicolas Le Scouarnec are respectively Principal and Research Engineers at Broadpeak. In this paper, they propose a novel scheme Server-Side Segment Selection for streaming (S4S), which allows players to interact with an S4S-enabled server to improve the overall experience.
Over-the-top (OTT) streaming growth is driven by the increasing number of wireless connected devices such as tablets and phones. An enabler for using these devices is adaptive bitrate (ABR) streaming that allows devices to select at each moment the stream quality that best fits the available bandwidth. The overall experience is highly dependent on the bandwidth estimation. The current approach to bandwidth estimation in HTTP-based ABR players is challenged by the evolution toward low-latency protocols (DASH CTE—Dynamic Adaptive Streaming over HTTP—Chunked Transfer Encoding) or (HLS LL—HTTP Live Streaming—Low-Latency), which result in micro-burst of traffic. Indeed, estimation at the client side, at the HTTP level, relies on the assumption that a relatively large segment of data is available and can be sent at the link speed, which is not the case anymore with low-latency protocols. To address this issue, we propose a novel scheme Server-Side Segment Selection for streaming (S4S), which allows players to interact with an S4S-enabled server to improve the overall experience. First, bandwidth estimation is done at the server side, using transport’s congestion control sender-side stats, leading to more precise estimates even in the presence of small burst, especially when using modern congestion control algorithms such as Bottleneck Bandwidth and Round-trip propagation time (BBR) or Performance-orientated Congestion Control (PCC). Second, we define a protocol for allowing the network to control the bandwidth versus quality tradeoff .
Where to read the full article?
How to cite?
G. Bichot and N. L. Scouarnec, “Server-Side Segment Selection for Low-Latency Streaming—S4S,” in SMPTE Motion Imaging Journal, vol. 130, no. 10, pp. 50-56, Nov.-Dec. 2021, doi: 10.5594/JMI.2021.3117125.