Optimal Resource Allocation in Distributed Multimedia Servers for Video-Audio Streams Access

摘要

In this paper, we present a theoretical model of resource contention problem encountered in conventional distributed multimedia servers and propose an optimal algorithm to solve the problem. Demands for multimedia streams retrieving from Internet impose a challenge problem on multimedia servers to deliver the real-time streams seamlessly and maximize system throughput. Especially, for a large scale of multimedia servers consisting of tens/hundreds of nodes interconnected with an interconnection network, the problem becomes more stringent. To serve requests of audio/video stream retrieval, proper system resource (e.g. computing power of the server node and the network bandwidth of the links alone the path to deliver the stream) should be allocated. Improper resource allocation would lead to hot-spot problems on the shared resources. It thus results in low system utilization and throughput. We consider heterogeneous types of requests in this paper. In addition, without lost of generality, we do not restrict topology of underlining interconnection network to any particular configuration. We model the problem as a networkflow problem and proposed a two-pass flow (TPF) algorithm to deal with the flow problem involving multiple flow types.