SplitQuest: controlled and exhaustive search in peer-to-peer networks

This paper presents SplitQuest, a controlled and exhaustive search protocol that relies on a hybrid network structure to avoid unnecessary query replication and to speed up query propagation in P2P networks. In SplitQuest, peers are organized in replication groups, in which each peer shares its contents with all members, and queries are propagated only once to a group. By avoiding query duplication, directing queries to disjoint groups, and exploiting peers’ heterogeneity, SplitQuest is able to achieve high levels of recalls and low response times, while incurring very low overhead. We simulate SplitQuest using synthetic and traces of real-world topologies and show that it outperforms the best known solution in number of messages, response time, number of hops, and success rate for query resolution, while being resilient to high churn rates. We also derive upper bounds on query routing for SplitQuest.