Proactive Edge Computing for Video Streaming: A Mutual Conversion Model for Varying Requirements on Representations

The vast proliferation of video streaming traffic imposed unprecedented challenges to origin servers in the Core Network (CN), further exacerbated by the diversity of requirements on video encoding format, resolution, and frame rate. Multi-access Edge Computing (MEC) has been introduced to mitigate the problem by storing and transcoding videos at the network edges to reduce traffic to CN. Traditional approaches hardly convert videos mutually or successively, limiting the overall performance of edge servers. By applying the space-time video super-resolution (STVSR) algorithm and transcoding on edge servers, this work proposes a novel mutual conversion model with a conversion graph among representations. The model measures and estimates the qualities of conversion outputs to maintain high Quality of Experience (QoE). Moreover, an off-peak proactive cache replacement algorithm is introduced to utilize remaining resources by prefetching and pre-converting popular representations, catering to the changing interests of clients. Transmission history is also collected to predict band-width and playback buffer length. The experimental results demonstrate that the proposed approach significantly reduces traffic to the origin server while achieving high QoE.