Research On Internet Topology And Key Technologies

With the development of social economy and new technologies, the Internet is growing at a scale previously unseen. A detailed network map is critical for evaluating the performance of network protocols, for assessing the effectiveness of proposed techniques to protect the network from nefarious intrusions and attacks; meanwhile it is the foundation of multiple path properties measurement. Due to privacy and security reasons, Internet Service Providers (ISPs) keep their router-level topology information confidential; this policy introduces a series of topology analysis algorithm. Most studies utilize the well-known Internet debugging tool traceroute to collect path traces from topologically diverse set of vantage points periodically and use this data to build a network map at different level. This thesis focuses on the route-level topology of domestic single ISP.This thesis studies the algorithms of IP alias resolution and presented an improved algorithm by proposing a new rule during the alias period, which was called Falut Tolerent Alias Resolution (FTAR). This method can filter wrong path traces due to certain types of load balance by sacrificing the alias completeness. Incremental measurement was adopted to improve coverage of the Internet. Alias completeness can be improved without imparting severe measurement traffic load on the network by several times of incremental measurement.Algorithms of route-level topology construction and prediction of the routes between any arbitrary pair of end-hosts in the ISP were studied and proposed. Each link in the collected topology was assigned with a weight value according to the shortest path criterion. The technique that was employed to predict routes between a pair of nodes was to compose partial segments of known Internet paths since the Internet predominantly uses destination based routing policy. The maximum like predicted path was selected by a model considers both network performance and the short path criterion if more than one predicted paths were observed.This thesis implements topology measurement and prediction of real ISP. Methos proposed above can meet the demand of actual network measurement and provided the basis for further research.