基于代理缓存的流媒体动态调度算法研究

Study on Dynamic Scheduling Algorithms for Streaming Media Based on Proxy Caching

该文研究了在 Internet环境下通过代理服务器将流媒体从源服务器传输到多个客户端时的骨干网络及服务器带宽消耗问题.在分析了传统多媒体流动态调度算法不足的基础上,提出了基于代理服务器协助的补丁预取与服务调度算法(P3S2A).由代理服务器通过单播连接从源服务器中获取流媒体数据,然后通过组播方式转发给客户端.同时根据当前客户请求到达的分布状况,代理服务器为后续到达的客户请求进行补丁预取及缓存.理论分析及实验结果表明,与传统的动态调度方法相比,该文提出的算法能够在不消耗更多的缓存空间的情况下,更有效地节省骨干网络的带宽资源,特别是在访问请求强度较高的情况下,效果更为突出.

This paper investigates the backbone network and server bandwidth consumption of efficiently delivering streaming media from a remote server through a proxy to multiple asynchronous clients in Internet-like environments, where the server-proxy network connection provides unicast-only service, or the multicast capability is only offered on the proxy-client path. Based on the previous work, the authors propose proxy-assisted patch prefetching and service scheduling algorithm-P3S2A, which tightly combines the transmission scheduling schemes with proxy caching. In the proposed scheme, the proxy retrieves the entire multimedia stream from server through the unicast connection, and then delivers it to the clients via multicast. At the same time, the proxy adaptively prefetches the patching data for the subsequent requests from the ongoing entire stream and caches them at the buffer, which just depends on the current batching interval that has non-zero requests. As compared with the classical dynamic streaming scheduling strategies, the proposed algorithm is shown that it can reduce significantly the consumption of aggregate bandwidth on backbone link within much wider range of request arrival rate, but not consume more story space at proxy.