Overlay network for low churn environment

Abstract

Distributed Hash Tables (DHTs) are the important algorithms for building structured peer-to-peer overlay networks. There are two approaches that used by the DHT's designers for choosing the number of states maintained by each peer. With the first approach, each peer stores minimal information about other peers, typically increase logarithmically with the number of peers in the network. The DHT protocols need to minimize the communication cost for updating the overlay states with the dynamics of peers that join and leave the network arbitrarily (also called as churn). Maintaining minimal states under churn condition will minimize network updates. With the second approach, each peer stores complete information about all peers in the network. Key lookup messages can be resolved by contacting one peer, thus decreasing the hop count for forwarding key lookup messages to single hop. The DHT protocol performance come with higher bandwidth consumption for updating routing states under chum condition. In this paper we evaluate and compare both approaches and show that the second approach performs better in low churn environment, thus can be used for application with less dynamic peers behavior.