A QoE-based caching algorithm for HTTP adaptive streaming contents in radio access networks

The double-digit growth of the number of mobile broadband subscribers has resulted in an exponential growth of mobile traffic in recent years. A major part of mobile traffic is the duplicate downloads of popular contents. A cache-enabled radio access network in which the popular contents are replicated at the eNodeBs (eNB) helps to improve the quality-of-experience (QoE) of the end users and to reduce the transit traffic on the backhaul links. Recently, adaptive streaming over HTTP has become a major technique to deliver the video contents to the end users in the Internet. With this technique, each video content is stored in several representations corresponding to different QoE levels. In this paper, we introduce an integer linear programming (ILP) model for video content caching which maximizes the sum of QoE supported by a cache-enabled eNB. An online caching algorithm is proposed based on the greedy algorithm solving the LP-relaxation of the ILP. The simulations show that our proposed algorithm yields a higher QoE, cache hit ratio, and a lower transit traffic on the backhaul links than the caching with traditional eviction algorithms such as least-recently-used and least-frequently-used.