QoE-driven Cache Placement for Adaptive Video Streaming: Minding the Viewport

Abstract

To handle the increasing demand for video streaming, ISP's and service providers use edge servers to cache video content to reduce the rush on their servers, balance the load between them and over the network, and smooth out the traffic variability. The dynamic adaptive streaming over HTTP protocol (DASH) makes videos available in multiple representations, and end-users can switch video resolution as a function of their network conditions and terminal display capacity (e.g., bandwidth, screen resolution). In this context, we study a viewportaware caching optimization problem for dynamic adaptive video streaming that appropriately considers the client viewport size and access speed, the join time, and the characteristics of videos. We formulate and study the proposed optimization problem as an Integer Linear Program (ILP) that balances minimal join time and maximal visual experience, subject to the cache storage capacity. Our framework sheds light on optimal caching performance. Our proposed heuristic provides guidelines on the videos, and the representations of each video, to cache based on the video popularity, its encoding information, and the distribution of end-user display capacity and access speed in a way to maximize the overall end-user QoE.