资源描述:
《network congestion arbitration and source problem prediction using neural networks》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、SmartEngineeringSystemDesign,vol.4:243-252,2002NETWORKCONGESTIONARBITRATIONANDSOURCEPROBLEMPREDICTIONUSINGNEURALNETWORKSJ.ALANBIVENSBOLESLAWK.SZYMANSKIComputerScienceComputerScienceRensselaerPolytechnicInstituteRensselaerPolytechnicInstituteTroy,NewYork12180-3590Tr
2、oy,NewYork12180-3590bivenj@cs.rpi.eduszymansk@cs.rpi.eduMARKJ.EMBRECHTSDecisionSciencesAndEngineeringSystemsRensselaerPolytechnicInstituteTroy,NewYork12180-3590EMBREM@RPI.EDUABSTRACTRapidlygrowingneedsfornetworkingintheInternetandinintranetsmakenetworkmanagementinc
3、reasinglyimportantintoday'scomputerworld.Weproposeusinglearningtechniquestopredictnetworkcongestionproblemsbeforetheystartimpactingtheperformanceofservices.Inthispaperwefocusonusingasimplefeedforwardneuralnetworktopredictseverecongestioninanetwork.Wealsouseneuralne
4、tworkstopredictthesourceorsourcesresponsibleforthecongestion,andwedesignandapplyasimplecontrolmethodforlimitingtherateoftheoffendingsourcessothatcongestioncanbeavoided.UnliketheusualTCP/IPflowcontrol,theproposedmethodisappliedonlytoselectednodesandconvergestothefin
5、alratefaster.Thedescribedtechniquessetthestageforanewwaveofnetworkmanagersthatarecapableofpreventingnetworkingproblemsinsteadofrepairingthem.KEYWORDS:congestioncontrol,neuralnetworks,weightpruning,networkproblemprediction,flowratecontrolINTRODUCTIONInthegrowingworl
6、dofnetworking,moreandmoreemphasisisbeingplacedonspeed,connectivity,andreliability.Networkperformanceisvitaltobusinessesforinter-businessoperationsaswellasbringingaproducttotheconsumersthroughelectroniccommerce.Networkinghasinfluencedoureverydaylives,astheinterconne
7、ctivityoffamiliesandfriendshaschangedthewaysinwhichtheycommunicateandseekinformation.Withthegrowingnumberofnetworkusers,moreemphasisisplacedonthemaintenanceandreliabilityofthenetworks.Whennetworkproblemsoccur,theyoftencatastrophicallybreaktheserviceforthoseenterpri
8、sesorindividualsthatdependonthenetworkconnection.Sometimes,suchbreaksofservicearejustannoying,butforcompaniesandcommercialuserstheyoften1meanlost