• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

principle|anodemaintainsonlyalimitedamountofinformationwithrespecttoaboundedset temssuchaspeer-to-peernetworks.Weproposeadenitionofself-organizationbasedonthelocality ofnearbynodes.Asopposedtotraditionalpeer-to-peersystemsthatfocusondataplacementand requestrouting,wepresentadualapproachwhichreorganizeslinksandspontaneouslygroupsnodes usingsimilaritycriteria,thusextendingthelocalityprincipletothenetworktopology.Moreoverweproposeanarchitecturalmodelofthepeer-to-peernetworksrelyingontheself-organizationofthesystem inmultiplelayers,eachofthemreectingtheself-organizationwithrespecttoaparticularevaluation criterion(e.g.connectivity,loadbalancing,datasimilarity...).Thisarchitectureisadaptedtothe 1Introduction trackingservice. implementationofascalablesetofservices.Wepropose,asanexample,animplementationofadata nodesarecontinuouslyjoiningandleavingthesystem,thereisnocentralentityinchargeoftheirorgani-Themajorfeatureofallrecentscalablesystemsistheirextremedynamisminstructure,contentandload[17]: self-organizationishandledthroughprotocolsfornodearrivalsanddeparturesbasedonafault-tolerant data[10].Tocopewithsuchcharacteristics,thesesystemshavetobeabletospontaneouslyorganizetoward desirableglobalproperties.Dierentkindsofself-organizationcanbeaddressed.Inpeer-to-peersystems, zationandcontrol,andthereisanequalcapability,andresponsibilityentrustedtoeachofthemtoown overlaynetworkasinPastry[7,16],orthroughlocalizationandroutinginfrastructureasinOceanStore[11,

Citations

No citations identified.