TITLE: Exploring a peer-to-peer protocol for streaming content AUTHORS: Abhinav Mishra & Gesly George AIM: We propose a P2P architecture for streaming content (primarily video). In the current architecture, with a single centralized server, as the number of clients increase, the overall performance at the server deteriorates. In a P2P setting, where each client must serve at least one other client, the overall performance should increase with an increasing number of clients. The current architecture has clients connecting to a single well-known server. In our proposed architecture, the well-known entity is the tracker and there is a single source server. The clients request streams by contacting the tracker which redirects them to another entity (either the source server or another client). This entity is now acts as the server for the new client and the new client obtains the stream from it. This then grows into a n-ary tree with the server at the root. The specific application where this can be used is real-time video. Examples include global webcasts and live streaming of sports events. ISSUES TO BE ADDRESSED: The obvious problem here is the issue of delay. Clients one step away from the root shall receive the stream after some delay (say, D). The clients two steps away from the root, shall receive the stream after some larger delay (roughly 2D). In general, the further away from the client is from the root, the larger the absolute delay it experiences. In most cases, however, it does not really matter how much time elapses between the event actually occurring and being seen at the client. The important thing is that all events be seen in the same order as they occur, all with the same delay. Nevertheless, we still have to ensure ordered delivery. In most cases, however, this absolute delay is unimportant. The critical property to be ensured is that all events be observed in the same order as they occur, all with the same delay. Nevertheless, we still have to ensure ordered delivery and there should be no observed jitter. EXPECTED RESULTS: We expect that this peer-to-peer architecture shall scale better than the existing single-server architecture.