A Comparative Study of Cache Replacement Algorithms for Video Streaming

Abstract

The increasing demand for video services has made video caching a necessity to decrease download times and reduce Internet traffic. In addition, it is very important to store the right content at the right time in caches to make effective use of caching. An informative decision has to be made as to which videos are to be evicted from the cache in case of cache saturation. Therefore, the best cache replacement algorithm is the algorithm which dynamically selects a suitable subset of videos for caching, and maximizes the cache hit ratio by attempting to cache the videos which are most likely to be referenced in the future. In this thesis we study the most popular cache replacement algorithms (OPT, CC, QC, LRU-2, LRU, LFU and FIFO) which are currently used in video caching. We use simulations to evaluate and compare these algorithms using video popularities that follow a Zipf distribution. We consider different cache sizes and video request rates. Our results show that the CC algorithm achieves the largest hit ratio and performs well even under small cache sizes. On the other hand, the FIFO algorithm has the smallest hit ratio among all algorithms.