Passive Measurement Based Network Topology Inference And Visualization

As the network structure becomes more complex,it's increasingly difficult for network administrators to manage network.The best way for them to analyze network traffic,locate the fault device and optimize network configuration is to obtain the net work topology through network topology inference and use visualization technology to display it.Passive measurement based network topology inference can be divided into specific protocol based network topology inference and Time-to-Life based network topology inference.The former requires the network equipment to support some specific protocols when infering network topology such as OSPF,and the weakness of this method is that it can't infer the network topology if the network equipment does not support specific protocol.The latter needs to use a small number of active measurements when infering the network topology,which faces two problems,the one is that it has low integrity of network topology and the other one is that this method cannot obtain net work topology if the active measurement is prohibited.In addition,how to display a complex network topology in a n intuitive way is also a big problem.In order to obtain network topology in the case of reducing the number of probe packets or without active measurement and display it in a clear way,we focus on Time-to-Life based network topology inference and visualization.First,we briefly introduce the background and research status of network topology inference and visualization.Some important protocols are also described.Then we study several network topology types.We describe the concept of visualization and compare several visual layout algorithms.Second,we propose a Time-to-Life based network topology inference algorithm which is named PFTTL and we can use it to infer the network topology by hop-count and the prefix matching length of two IP address without sending any probe packets.On the basis of PFTTL,we analyze NDTTL which is also a Time-to-Life based network topology inference algorithm and point out that NDTTL has the problem of low integrity.Therefore,we propose EPREDICT to solve the problem by predicting the number of edges.EPREDICT was tested on data and the result show that EPREDICT has higher integrity than NDTTL.Finally,on the analysis of user 's requirements,we design and develop a topology visualization system.The system has six modules,including data reading module,query statistics module,inference algorithm module,network layout module and network analysis module.We adopt B/S structure to develop topology visualization system and use My SQ L to store data.In addition,we use HTML5,CSS and JQuery to develop the user interface of the system and use ECharts to display figures.Topology visualization system was tested and the result show that the system fulfill user's requirements.