Evaluation Of Network Survivability Of The City-level Internet Topology Under Regional Failures

As a national information infrastructure, the Internet has affected all aspects of social life deeply. In order to provide uninterrupted network services to users, network survivability turns to be a very important and challenging subject. From the 2003 North American blackout and the 2006 Taiwan earthquake, we can see that regional failures on the Internet may have cascading effects, resulting in a network paralysis of a large area. These regional failures exhibit high correlation among geographical space. Therefore, research on network survivability of the city-level Internet under regional failures is of great theoretical and practical value.In this paper, three aspects of network survivability of the city-level Internet topology under regional failures are studied.(1)We analyze the network survivability under two different regional failure models. We model the network with geographical locations of every city; then according to different failure inducements, we propose binary and probabilistic regional failure models; and several metrics are defined to measure the survivability of the network. After that, we analyze survivability of China city topology and that of US city topology under two regional failure models through simulations. The results show different impact of failure models and different degree of survivability of China and US networks. We also get the geographical importance of every city.(2)Based on the OpenMP and Parallel Studio software, we improve the single core program for computing the basic properties of network topology to a parallel program. Experimental results show that, with the help of a Shared parallel programming mode, the capability of multi-core is developed sufficiently, to get the ideal speedup and reduce the computing time greatly.(3) We have designed a software tools for network survivability evaluation. Based on the topology files and city location data, this tool automatically generates maps of network topology with location information and distributions of geographical characteristics, with calculation of the basic properties of network topology. According to user's selection, it automatically evaluates the city's geographical importance. It can also analyze the relationship of failure radius and survivability under dualistic regional failure model as well as the relationship of failure impact and survivability under probabilistic regional failure model, with a clear visual representation of the evaluation results.This work provides a method to study the survivability of large-scale complex network under regional failures, and has theoretical and practical significance of improving researches and applications on this subject. In future work, we will continue to develop more accurate failure models and refine the geographical topology to make the research closer to reality.