Design And Implementation Of Network Probe Software Architecture

Measurement is one of the key approaches to acquire the network performance.WiththeevolutionofInternet,newapplications,suchasstreamingmedia,VoIP,videoconference, P2Petc, become more and more popular. Monitoring and accurate trafficidentification of new applications can shed light on traffic characteristics and itslong-term trend, and it is the base of many other network activities, from trafficshaping,trafficengineeringtoeventualtrafficcontrollingandQoS.This paper analyzesthe design of recent network measurement infrastructure andimplements new network measurement probe architecture. To overcome theshortcoming in traffic identification capability in many network measurementinfrastructures, we study the traffic characteristic of multimedia applications. Andcorresponding traffic identification methods and an identification framework areproposed.Itisnowintegratedintothenetworkmeasurementprobe.Themaincontributionsofthisthesisareasfollows:(1) Network probe system is designed and implemented. It is composed ofmeasurement platform and measurement tools. The former is independentfrom the measurement tools. Plug-in measurement tools make the probe quitescalable. The probe supports active and passive measurement. Triggeredmeasurement scheduling and time deviation revision algorithm make thescheduling of measurement tasks much more accurate than before. Built-infirewall and application authorization mechanism prevent unauthorizedconnection.(2) Identification methods of new applications are studied. The signature stringpattern of mainstream multimedia control protocol is defined. With protocolpayload analysis, multimedia traffic based on RTSP, MMS, SIP and H.323 isidentified.P2P-based online streaming media traffic generated by PPLive andPPStreamcanbeidentifiedbasedonthesignaturestring.(3) A scalable traffic identification framework is proposed. It uses port-basedmapping,signature-basedidentification andpayload-based analysis toidentifytraffic of different applications. Identification of many applications traffic issupported and can be further extended. It is now implemented and integratedintothe probesystem.