Anycast-based context-aware server selection strategy for VoD services

Abstract

Internet video services have experienced an immense popularity in the recent years and will probably be the dominant applications of Future Internet. Therefore, the design of scalable video distribution systems that on the one hand, improve the end-users experience and deal with theirs context heterogeneity and on the other hand, optimize the cost of these services, is becoming a very important and challenging issue. One common approach for this purpose is the server replication. Nevertheless, this approach brings new issues such as the server selection (how to direct the user request to the closest replica), the servers' placement, etc. This paper deals with the server selection issue. Thus, we present a two-level filtering based server selection. The policy-based filtering retrieves the list of servers whose services matches the client context while The metric-based filtering select from this list the most convenient server that on the one hand, avoid the servers overload and in the other hand, copes with the underlying network state and capabilities. The paper also presents the preliminary evaluation of the proposed video provisioning scheme. The promising results led to more extensive evaluation of the approach and later to the launch of real developments in the framework of a large-scale European project which objective focuses on Future Media Internet.