《Optimizing Bloom Filter Settings for Multi-keyword Search in Kademlia-like DHT Peer-to-Peer Networks》.pdf

ChineseJournalofElectronics Vo1．23，No．2，Apr．2014 Optim izingBloom FilterSettingsfor M ulti-keyword Search in Kadem lia—likeDHT Peer—to—PeerN etworks水 MAWenming一．ZHANGYujie，andMENGXiangwu， (1．SchoolofComputerScience，BeringUniversity PostsandTelecommunications，Bering100876,China) (2．BeijingKeyLaboratoryofIntelligentTelecommunicationsSoftwareandMultimedia，Beifing100876，China) A bs·tract— ThisPaPerhasproposed a novelm ethod tothepeerresponsibleforholdingthecorrespondingreverse toreducethem ulti．keywordauery trafflein Kadem lia-like index．Partition—by-keywordusesDHT tostoretheindicesof Peer-to-peer(P2P)networksbyoptimizingtheBloom lil- files．andusesthesetablestofindthelpcationofatargetdoc— tersettings．W elbuild som e m odelsto estim ate the corn ． ument．Ithashesthekeyword ofadocumentand storesthe m unicationcost，theunionsetsize，andthelossrateofPer— indexofthedocumentinanodewhichcontrolsthehashtable form ing union and intersection operations．W _eim plem ent a Kadem lia-like system and generate a group ofdatasets． c0rrespondingtothekeyword’shashvalue．Sinceadocument 、^，_e use one part ofthe datasets to obtain the functions can containseveralkeywords，theindexofthisdocumentcan how to com pute the optim alparam etersand use another bestored inanumberofnodes．Tofindthelocation ofadoc partofdatasetsto verify ourm ethod．Each query can de- umentthatcontainsaspecifickeyword，theclientpeersimply term inetheoptim alsettingsofBloom filterwith no extra hashesitandsendsaquerytoanodewhichcontrolsthecor— configuration． Our sim ulation experim entalresultsshow