Anti-jitter P2p Overlay Network Design And Analysis

Churn in Peer-to-Peer (P2P) networks is the phenomenon that a large number of nodes join, leave or fail spontaneously and frequently, because of their nature of anonymous, free, and large-scale. Churn is such a significant performance influential factor of P2P networks that, it may cause date loss, search failure, inconsistency search results, increase in user-related latency and bandwidth, or even partition of the overlay network, which will lead to sharp system performance decrease or even that the overlay can never recover. Besides, churn is also one of the most important reason why structured P2P network models that have better performance expectations than unstructured ones have not been deployed by now. Hence, whether in designing a new P2P protocol or when a P2P protocol is evaluated, the ability to recover from churn should be considered. Therefore, it is very meaningful to study the characters and nature of churn in P2P overlay networks, and try to find schemes to make P2P systems more churn-resistant.In this paper, we first sum up the state of art of study in nature of churn in deployed P2P networks and in churn-resilient strategies, and then build up two churn-resilient P2P networks and an algorithm, as follows:(1) A novel structured P2P network of high churn-resistance-CT-Cycloid, which is based on Cycloid, is constructed by use of redundancy mechanisms, lifespan-based node choice mechanisms, and a role-division scheme.(2) A new topology-aware Kelips overlay network is constructed by exploiting and utilizing topology proximity through "landmark clustering", and by modifying the original gossip protocol. This network performs periodic recovery through a gossip recovery protocol, and has higher efficiency and stronger ability of churn-resilience.(3) A load-balancing algorithm which can be used to build churn-resilient P2P overlay networks is designed. The algorithm is based on virtual server, and structured P2P networks with load balancing and high churn-resistant capability can be built in the process of the algorithm's implementation.