PDStream: Slashing Long-Tail Delay in Interactive Video Streaming via Pseudo-Dual Streaming

TL;DR

PDStream tackles the long-tail end-to-end delay in interactive video streaming caused by bursty keyframes. It introduces a pseudo-dual streaming mechanism that activates a parallel non-keyframe playback stream while the primary stream preserves keyframes for reference, guided by a frame-level R-Q based bitrate model and RL-driven adaptation, with priority-based transmission to minimize playback delay. Across a WebRTC IVS testbed and real-world traces, PDStream achieves a 17.5% reduction in average E2E delay and a 33.3% reduction in the 97th percentile, while maintaining or improving video clarity under dynamic bandwidth. The work demonstrates the practicality and effectiveness of decoupling playback and reference functions via lightweight dual streaming, offering broad applicability to WebRTC-based IVS applications without codec modifications.

Abstract

End-to-end (E2E) delay is critical for interactive video streaming (IVS) experiences, but remains unsatisfactory for its long-tail distribution caused by periodic large keyframes. Conventional optimization strategies, such as jitter buffer, bitrate adaptation, and customized encoding, either sacrifice clarity, average delay, or compatibility. To address this issue, we propose PDStream, a novel pseudo-dual streaming algorithm, aimed at minimizing E2E delay while maintaining video clarity. The core idea is to split the two functions, delay-sensitive playback and delay-tolerant reference, on keyframes through dual streaming. Specifically, the playback function is held by a second parallel stream, which comprises much smaller non-keyframes and is allocated more immediate bandwidth for real-time performance. The reference function is ensured by the first stream with keyframe preservation, allocated more subsequent bandwidth to smooth out bursty traffic. Additionally, ``pseudo'' minimizes computational and transmission overheads by restricting dual streams to brief activation only when keyframes appear, supported by corresponding dual-stream bitrate allocation and adaptation to ensure delay and clarity. We implement PDStream on a WebRTC-based IVS testbed with real-world network traces. Results show that PDStream significantly outperforms prior algorithms, reducing average E2E delay by 17.5\% and slashing its 97th percentile by 33.3\%, while keeping clarity under varying bandwidth.

Figures (16)

• Figure 1: Keyframes are 5-10$\times$ larger than non-keyframes.
• Figure 2: Delay components under varying video bitrates and types.
• Figure 3: A showcase of E2E delay components.
• Figure 4: Long-tail distribution without bitrate adaptation.
• Figure 5: Long-tail distribution with bitrate adaptation.