Enhancing User Experience and Storage Efficiency for Adaptive Video Streaming in Edge Networks

The exponential growth in video traffic presents substantial challenges to the existing content delivery networks. Mobile Edge Computing (MEC) emerges as a viable solution to quickly and directly respond to user requests by caching video sources in close proximity to users. However, existing works primarily focus on caching strategies with fixed bitrate policy, ignoring the interaction between cache deployment and bitrate decision and leading to a waste of edge cache capacity. In this paper, inspired by the great success of neural networks (NN) based codec, we demonstrate that the introduction of NN-based codec can achieve lower cache storage with additional decoding time cost. Accordingly, we formulate a min-max optimization problem for the joint edge caching strategy and multi-codec bitrate decisions, aiming to optimize cache storage utilization while providing high-quality user experience. To cope with the NP-hard nature of min-max problem, we decompose the original problem into a two-stage joint iterative optimization that includes multi-codec bitrate selection subproblem and chunk cache deployment subproblem. In particular, we design an origin-centric video response paradigm to optimize multi-codec bitrate selection stage. While ensuring user Quality of Experience (QoE), the proposed paradigm downloads chunks as much as possible from the origin, thereby differentiate the edge usage of different chunks. Based on the edge usage of different chunks, we can derive the cache deployment by solving the cache utility maximization problem based on sampled edge request rate and chunk size. We conduct extensive experiments utilizing real-world network trace datasets and video files. Compared with representative benchmarks, the proposed method significantly reduces edge cache storage by 8.99% to 57.96% while providing 0.17% to 0.36% extra QoE.