Behavior-aware p2p protocol for high data availability

Abstract

A lot of research papers discussed the Distributed Hash Table (DHT) based p2p systems to promise that idle resources may be efficiently harvested. However, p2p systems are composed of components with extremely heterogeneous availabilities and to handle churn, the system will generate the heavy information maintenance traffic to keep the efficiency of the DHT. Moreover, the previous researches concentrate on the efficient lookup of the p2p system and they fails to keep high data availability effectively. This dissertation presents the behavior-aware p2p protocol for high data availability. By using this protocol, the p2p system can reduce the overhead by exploiting the heterogeneous behavior of participant nodes efficiently. Unlike the DHT based p2p which use the static nodeID which is obtained by wellbalanced hashing function, the behavior-aware p2p ignores this static nodeID and uses the dynamic nodeID. This dynamic nodeID is composed of the Load Balanced ID which balance the loads of the reliable and powerful nodes and the Load Free ID which reduce the responsibility of normal and weak nodes and eliminate the compulsory maintenance overhead when the churn of them occurs. The assignment of nodeID performs without any central management server. The new nodeID is assigned according to the current network state and the node which processes the join message. After assigning the new nodeID, this nodeID of a node changes on the fly according to its behavior on the p2p system. Finally, every node gets its proper nodeID with its characteristics and each nodes takes the different responsibility in accordance with its nodeID to support p2p system effectively. The reliable and powerful nodes mainly guarantee the data availability and the correctness of lookup. The normal and weak nodes assist them and the churn of them can not harm the data availability and the efficiency of the lookup. I examine the efficiency of behavior-aware p2p via a event driven simulation...