摘要
Thisletterreportsastudyofahybridburstassemblyandahybridburstlossrecoveryscheme(delay-basedburstassemblyandhybridlossrecovery(DBAHLR))whichselectivelyemploysproactiveorreactivelossrecoverytechniquesdependingontheclassificationoftrafficintoshorttermandlongterm,respectively.TrafficpredictionandsegregationofopticalburstswitchingnetworkflowsintothelongtermandshorttermareconductedbasedonpredictedlinkholdingtimesusingthehiddenMarkovmodel(HMM).ThehybridburstassemblyimplementedinDBAHLRusesaconsecutiveaverage-basedburstassemblytohandlejitterreductionnecessaryinreal-timeapplications,withvariationsinburstsizesduetothenon-monotonicnatureoftheaveragedelayhandledbyadditionalburstlengththresholding.ThisdynamichybridapproachbasedonHMMpredictionprovidesoverallalowerblockingprobabilityanddelayandmorethroughputwhencomparedwithforwardsegmentredundancymechanismorpurelyHMMprediction-basedadaptiveburstsizingandwavelengthallocation(HMM-TP).
出版日期
2014年04月14日(中国期刊网平台首次上网日期,不代表论文的发表时间)