Network Performance Measurement And Speculate Technology

Internet is the biggest distributive network in the world. With the rapid developments, Internet became more complex than ever. It is a challenge to manage a network like this. The most challengeable thing may be the measurement of Internet. Because of its distributive, no corporative and heterogeneous characteristic, accurate measurement of such a network is very difficult.A lot of tools for Internet measurement have been developed, such as ping, traceroute and etc. But most of them require the cooperation with network nodes, or deploying the measure node in the network. The measurement may affect the network traffic, and cause the security problem. A promising technique is network tomography, which infers the inner network parameters by end-to- end measurements. The technology is similar as the computer tomography (CT). It neither needs to deploy a measurement node inside the network nor requires a cooperative node.In this thesis, the network tomography algorithms are investigated. The link's packet loss ratio and delay distribution are inferred using these algorithms by end-to-end measurements. The work of this thesis consists of three parts. Firstly, literal review and comparison of these algorithms are done. Secondly, some new algorithms are proposed. A new algorithm based on the multicast dependence tree model is proposed to infer the inside link's packet loss ratio and delay. The Marcovian character of the tree is naturally incorporated to reduce the computing complexity. Simulations are done to show the effectiveness of the algorithm. Another improved algorithm based on the R.Nowak's work is proposed to make it more suitable for online calculation. In order to infer the continuous probability distribution, an improved algorithm is proposed by incorporating auxiliary information in Mixed Finite Mixture model. The precision of the estimated parameters is improved. Finally, a network measurement framework is proposed in the thesis by using network tomography technology. The mature tools of network measurement haven't adopted tomography technology in the world. A system based on the framework had been constructed. It is convenient for different algorithms to upload/download dynamically. It is flexible to extend the system.