N.V. Pavlenko

Recently the IP-TV systems whose primary purpose is to deliver a continuous media stream to a large number of users over a global computer network (Internet) have been getting more wide distribution. The analysis of the well known technologies in this area shows that for a such large scale broadcasting a peer-to-peer approach is the most promising technique. But due to significantly long connections in the Internet the most crucial difficulty when building such peer-to-peer systems is selection of a technique for media stream distribution among users. For all that the primary criteria of the efficiency assessment of these techniques is the average bitrate of the media stream being received by users as well as the steadiness of its arrival. In this paper a new method to build a peer-to-peer media stream distribution system based on the information about network topology is proposed. Such information allows to localize the process of media traffic exchange in local networks owned by large Internet providers by means of combining into groups the network nodes which are close to each other from the network topology standpoint. Such approach allows network nodes of a peer-to-peer system to exchange media traffic primarily with their neighbours in local networks and to less extent rely on media traffic from the nodes which are located far away for instance in other cities or countries. The application of a hierarchical cluster analysis to combine nodes into such groups allowed to significantly (1.5 - 2 times) accelerate the media stream distribution process which was proved during the conducted simulation. The properties of the simulation process are also shown in the paper