网络流量监测算法的的分析研究及其实现.pdf

南京邮电大学硕士研究生学位论文摘要摘要网络技术是一项劳动密集型产业,在比较复杂的网络环境中,不对网络流量进行实时的监控和适当的管理,很容易造成网络的繁忙,导致网络的拥塞等状况,影响网络用户的使用,直接导致经济的损失。为缓解网络繁忙这一压力,国内外诸多大型网络设备制造商一直在努力,寻求最佳的解决方案。目前现有的流量监测技术对流量监测主机的CPU和内存的要求很高,并且不适用于多条链路、大规模区域网络的集中管理和流量监测:监测代理的CPU负荷和内存负荷比较大; 输出信息不完善,不能得到关于网络接口的统计信息;统计信息输出过于频繁,在信息收集器和监测设备之间会占用较大的网络带宽。针对这些问题,Stream,该技术有效改善了现有流量监测技术的这些缺点。 NetStream的关键算法是其报文抽样算法,很多理论的随机抽样算法由于受硬件条件的限制无法实现,Stream的丌发条件和所用的硬件,设计并实现了快速伪随机抽样算法。在不影响数据统计的前提下,更有效的减轻NP对报文镜像所带来的负担,从实验结果分析,该算法是有效的。关键词:流量监测,流量采集,stream 南京邮电大学硕士研究生学位论文 Abstract work technology is a labor-intensive real-timemonitoring and appropriate management work work willbeeasilybusy,it can work congestion and ,users are affected and directeconomic losses OCCUr. Inorder toalleviate thepressure,a number workequipment manufacturers both at home and abroad have been trying hard tofindthebestsolution. Atpresent,existing trafficmonitoring technologies have highdemand on CPU andmemory of host for traffic monitoring,and do not facilitate centralized management and traffic monitoring on many linksandlarge—scale CPU loadandmemory load of Monitoring agent isrelatively large,the output information is not work interface statisticalinformation cannot beobtained。Information outputted is too frequent work band width between theinformation collector and themonitoring ,we design and realize trafficmonitoring Stream,which can improve some ings ofexistingtrafficmonitoringtechnology effectively. Stream’S essential algorithm is itspackage sampling lotofacademic random samplingalgorithms Can not beachieved due tohardware binding the development Stream and hardware,we design afastpseudo—random samplingalgorithm and realize algorithm ismore effective tomitigate NP’Sburden on thepacket mirroring without affecting thedata ofexperimental result prove thatthe algorithm iseffective. Keywords:Traffic Monitoring,Traffic Collection,Random Stream 南京邮电大学学位论文独创性