Guaranteeing the continuous stream service in cluster media servers

Abstract

Due to the explosive growth of the Internet, stream services are becoming more popular. The large media servers supporting thousands of concurrent media streams have to satisfy us with both good throughout and high availability. Although they are reciprocal from the viewpoint of system performance, the stream service continuity, as well as throughput, should be considered in clustered media servers. This is because media stream services should be continuously available to the clients even on the event of system failure. In this paper, we propose LSS and RRD mechanisms for the purpose of guaranteeing continuous stream services. These mechanisms provide additional functionalities for the traditional media servers. This paper focuses on how to apply these mechanisms to clustered servers with the minimum overhead according to the cluster size.