网络拓扑属性关联分析模型分析-analysis of network topology attribute association analysis model.docxVIP

哈尔滨工业大学工程硕士学位论文 哈尔滨工业大学工程硕士学位论文 - - II - Abstract With the advent of information era, the scale of Internet is growing at an exponential rate. At the same time, due to randomness, dynamicity and isomerism in itself, Internet becomes more and more complex. Therefore, the research method of large-scale network topology has attracted a growing interest in recent years. By researching the network topological data and the relationship among the attributes of network topology, the internal law of network topology and the basis for the evaluation of network topology can be obtained. This paper developed a research about the analysis model relevant to the attribute of network topology. Firstly, after analyzing the advantages and disadvantages of existing network topology detection method, this paper designs and implements an Internet network topology detection subsystem which is based on the Traceroute principle. The subsystem can probe Internet topology using TCP protocol as well as UDP protocol. By adopting tree detection and order detection and utilizing the reference path, the network load is decreased, and the performance of network t opology probe is improved. Then, in view of the correlation of topology attributes, this paper use Spearman method to divide the attribute of network topology into network scale and network connection class. According to the influence degree of each attribute in the appropriate category, this paper use AHP method to determine the weight of every factor and build an analysis model to guide specific network application. In the end, this paper designs and implements the network topological attributes associated with analysis system which combines the Internet network topology detection subsystem and network topology analysis model. The analysis of experimental data shows that the proposed probe method performs better than Traceroute in speed and integrity of probe. This paper also tests the rationality of relevant analys