欢迎来到天天文库
浏览记录
ID:37659434
大小:120.89 KB
页数:8页
时间:2019-05-27
《The Value of Consensus in Online Stochastic Scheduling》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、TheValueofConsensusinOnlineStochasticSchedulingRussellBentandPascalVanHentenryckBrownUniversityProvidence,RI02912frbent,pvhg@cs.brown.eduAbstractpracticalissuesarisinginonlineoptimization.Indeed,themainpurposeof(Chang,Givan,&Chong2000)instudy-Thispaperreconsidersonlinepacketschedulingincom
2、puteringthisproblemwastodemonstratethatexploitingstochas-networks,wherethegoalistominimizeweightedpacketlossticinformationmayprovidesigni®cantbene®tsinonlineandwherethearrivaldistributionsofpackets,orapproxi-packetscheduling.Theyproposedanexpectationapproachmationsthereof,areavailableforsa
3、mpling.Earlierworkproposedanexpectationapproach,whichchoosesthenextwhichchoosesthenextpackettoschedulebyapproximatingpackettoschedulebyapproximatingtheexpectedlossoftheexpectedlossofeachpacketclassoverthedistribution.eachdecisionoverasetofscenarios.Theexpectationap-Theapproximationsamplest
4、hedistributionstobuildscenar-proachwasshowntosignificantlyoutperformtraditionalap-iosfromavailableandfuturepackets.Theexpectationalgo-proachesignoringstochasticinformation.rithmwasshowntosigni®cantlyoutperformtraditionalap-Thispaperproposesanovelstochasticapproachforonlineproacheswhichignor
5、estochasticinformationsuchasgreedypacketscheduling,whosekeyideaistoselectthenextpacketheuristicsandalgorithmsoptimizingthedecisionsbasedonastheonewhichisscheduledfirstmostoftenintheoptimalknownpackets.However,theexpectationalgorithmmaybesolutionsofthescenarios.Thisconsensusapproachisshownqu
6、itedemandingcomputationally,sinceitsolveseachsce-tooutperformtheexpectationapproachsignificantlywhen-narioforeachclassofpackets.Asaconsequence,itmayevertimeconstraintsandtheproblemfeatureslimitthenum-notscalewellwhenthenumberofclassesincreasesanditberofscenariosthatcanbesolvedbeforemakingad
7、ecision.maynotbeapplicablewhendecisionsmustbetakenwithinMoreimportantlyperhaps,thepapershowsthattheconsen-susandexpectationapproachescanbeintegratedtocombinestrongtimeconstraints.thebenefitsofbothapproaches.Thispaperreconsidersonlinepacketschedulingandpro-These
此文档下载收益归作者所有