Multiple Small-Scale Floods in Unstructured P2P Systems

Gnutella-like unstructured systems usually use flooding with a high number of TTL hops to probe as many peers as possible within a short period of time. However, this mechanism is very inefficient because it handles all queries identically no matter how popular the querying objects are, hence results in a great amount of unnecessary messages. Aiming at improving search efficiency, we propose MSSF, an efficient search mechanism using multiple small-scale floods with different flood sources for each query. Based on MSSF mechanism, three policies are introduced in determining the order of floods initiated: sync, sequence and parallel. The sync policy tries to effect the response time for a query, the sequence policy is designed to minimize the number of messages by seeking to probe the minimum number of peers necessary to obtain the desired number of results, and the parallel policy considers a trade-off between response time and query traffic