Content subscribing mechanism in P2P streaming based on gamma distribution prediction

(整期优先)网络出版时间:2007-12-22
/ 1
P2Psystemsarecategorizedintotree-basedandmesh-basedsystemsaccordingtotheirtopologies.Mesh-basedsystemsareconsideredmoresuitableforlarge-scaleInternetapplications,butrequireoptimizationonlatencyissue.Thispaperproposesacontentsubscribingmechanism(CSM)toeliminateunnecessarytimedelaysduringdatarelaying.Anodecansendcontentdatatoitsneighborsassoonasitreceivesthedatasegment.Noadditionaltimeistakenduringtheinteractivestagespriortodatasegmenttransmissionofstreamingcontent.CSMconsistsofthreesteps.First,everynoderecordsitshistoricalsegmentslatency,andadoptsgammadistribution,whichpossessespowerfulexpressionability,toexpresslatencystatistics.Second,anodepredictssubscribingsuccessratioofeveryneighborbycomparingthegammadistributionparametersofthenodeanditsneighborsbeforeselectinganeighbornodetosubscribeadatasegment.Theabovestepswouldnotincreaselatencyastheyareexecutedbeforethedatasegmentsarereadyattheneighbornodes.Finally,thenode,whichwassubscribedto,sendsthesubscribeddatasegmenttothesubscriberimmediatelywhenithasthedatasegment.ExperimentsshowthatCSMsignificantlyreducesthecontentdatatransmissionlatency.