A Cluster-based Streaming Caching Proxy Framework

As streaming media over the Internet becomes popular, caching proxy has become increasingly important. Nowadays, caching proxy has turned into a standard technique, which is widely deployed on the Internet to reduce client-perceived delay and network traffic. In this paper, we propose a cluster-based streaming caching proxy architecture in which the local network has a number of independent caching proxy servers and a front-end machine called Director. The caching proxies are all coordinated and managed by the Director. Additionally, we present a novel cache replacement algorithm, called SPRA (Synthetic Popularity Replacement Algorithm) and explain the design philosophy. Simulation results demonstrate that our system achieves good scalability and the new algorithm outperforms many other popular replacement algorithms such as LFU and LRU-k with respects of byte hit rate (BHR), replacement times and initial latency. Key-Words: Streaming media, Caching proxy, Cluster, Meta information, Replacement policy, Byte hit rate