Abstract: In distributed streaming media delivery system, an important topic is how to reduce the consumption of backbone network bandwidth. By deploying the proxy at the edge of network, it made response to a group of local clients’ request and forwarded the multimedia data streamed out of the server. An efficient algorithm called batched patch pre-caching (BPP) was proposed, which tightly combined the server scheduling with proxy caching strategies. In terms of the distribution of clients’ request arrival at current time, the proxy allocated the corresponding buffer units in advance and cached data in segments from the ongoing multicast stream started at server. These cached data could be shared by subsequent requests till they were evicted from the proxy. The results of theoretical analysis and experimental simulation show that this algorithm can significantly reduce the bandwidth consumption on backbone link, and has very good adaptability for the variation of request arrival rate.