Study On Topology Inference Method Based On NT Technology Double-parameters

With the growth of the network’s size and complexity, the demands on networkmeasurement is also increasing, this greatly promoted the research and development ofnetwork measurement technology. Due to many limitations owned by traditional networkmeasurement methods (mainly performance in the traditional network measurement dependson the internal nodes’ collaboration in autonomous systems and the specific networkprotocols), a new network measurement technique network tomography (NT) was proposed inthe academia, which thinking comes from the idea of medical computed tomography scan(Computed Tomography). The main features of the NT technology are through the active orpassive methods based on end-to-end measurement to infer the various performancecharacteristics of the networks, without the needs of internal nodes’ collaboration. Networktopology inference is one of the major applications of network tomography technology, whichaccording to the correlation of nodes’ performance characteristics on the shared link to judgethe nodes’ connection relationship, and reconstruct the topologies of the network.This paper introduces the backgrounds and current situation of the research issue abouttopology inference technology which based on double-parameters in network tomography,and describes the related knowledge of network inference and network tomography techniquein detailed. Starting from the basis of topology inference and measurement techniques, thepaper Analyzes the existing topology inference method based on NT. Finally, two kinds ofdouble-parameters’ topology inference methods based on NT technique were proposed,simulated them in NS2to validate their effectiveness. The specific works schedules asfollows:First, introducing the issue’s research background and situation, describing the networkmeasurement’s basic principle, performance targets, classification, and the NT’s thesignificance, key assumption, main application, key techniques.Then, from the basic of topology inference and measurement methods, analysis theexisting topology inference algorithms based on NT. The most of the existing methods usesunique performance parameter to compute the between two nodes, so such methods waseasily affected by network background traffic. In order avoid those shortages, the topologyinference algorithm combined package loss rate and delay simultaneously is proposed, whichbased on immediately group measurement. In order to make the inferred logical topologymore similar to the realistic network topology, a sandwich packet sequence measurementmethod was proposed, the method describes an algorithm which combined the queuing delayand the time to live value to compute the correlations. Finally, with the help of the NS2simulation software, we established the network simulationplatform, performed simulation experiments of the proposed two methods, and uses the treeedit distance to analyzing, comparing and validating their effectiveness.