PlanetP : UsingGossipingandRandomReplication to Support Reliable Peer-to-PeerContent SearchandRetrieval �

We introduce the PlanetP system,which explores the construction of a reliable peer-to-peer (P2P) content search andretrieval serviceusingrandomlycirculatedglobal statebetweenpeers of an unstructured community. Our work representsa novel alternativeapproach to recentP2P systemsthat focuson enabling very largescalename-basedobjectlocation usingsophisticateddistributed data structures. We showthat our simpler approach scalesto several thousandpeers (ultimately targetingtheregimeof about tenthousand) andconvergesin several minutesusing only modest bandwidth while still maintaining reliable search, ranking, and retrieval similar to an Internet search engine. Unlike current search enginesor otherP2Psystems, however, PlanetPdoesnot require centralizeddirectoriesor management,nor builds a complex distributeddatastructure. PlanetPachievesits goalsusingthreemajor components.First,peers collaborateto maintainlocal copiesof theglobal member shipdirectoryalong with compact summariesof shared content usinga randomizedgossipingalgorithm. Second, peers implementsa per query, text-basedranking algorithm to helpusers ignore irrelevantdocuments.Finally, peerscollaborate to replicateunpopular content—popular content is naturally highly replicatedvia hoarding—usingReedSolomon erasurecodingto increasetheprobability of suc-