Joint Video Transcoding and Representation Selection for Edge-Assisted Multi-party Video Conferencing

Abstract

Current cloud-based multi-party video conferencing suffers from heavy workloads on media servers caused by video transcoding. Emerging edge computing can assist in offloading transcoding tasks to edge nodes. However, the resource-limited nature of edge nodes poses new challenges. First, edge nodes can real-timely transcode a video into only a subset of representations, raising the video transcoding problem of what is the set of representations each participant should transcode its video stream into. Second, since participants’ downlink resources are limited, one needs to solve the representation selection problem of what representation each participant should select for receiving another participant’s video. Third, the above two problems are coupled and should be optimized simultaneously. Hence, this paper studies the joint video transcoding and representation selection problem for edge-assisted multi-party video conferencing, with the aim of maximizing the overall QoE under the resource and real-time video transcoding constraints. Such a problem is formulated as a non-linear integer program and is NP-hard. To solve it, we leverage the submodular optimization technique and propose a $(1-\frac{1}{e})$ -approximate algorithm with the polynomial computation complexity. Finally, extensive trace-driven simulations are conducted to evaluate the proposed algorithm. The results show that it outperforms the alternatives by 1.5–2.5$\times$ on average in terms of overall QoE.