Research On Topology Aggregation Algorithm For Large-scale Network Simulation

With the rapid development of technology, network plays an increasinglyimportant role in our daily life. The network research of services for human beingshas drawn considerable interest in the academic community. In computer networkresearch, network simulation is a well-known technique with good accuracy andflexibility. Single network simulator cannot meet the research requirements oflarge-scale network simulation, and thus the parallel network simulation techniqueand network topology aggregative technique have been developed for large-scalenetwork simulation. However, parallel network simulation technique is with poorscalability, and its performance depends on some important factors, such asnetwork topology partitioning, remote routing strategy, multi-machinecommunication and load balancing. Therefore, there are some limits for large-scalenetwork simulation by using parallel network simulation. In this paper, weresearched network topology aggregative technique to improve the topologyaggregative level, reduce the scale of network topology, guarantee simulationaccuracy, and reduce the resource consumption of network simulation.Network topology aggregative technique focuses on topology aggregativelevel and simulation accuracy, and they are closely related to each other. In general,the higher the topology aggregative level is, the worse the simulation accuracy willbecome. Therefore, how to balance topology aggregative level and simulationaccuracy to achieve the best optimal performance and how to carry out topologyaggregation with different aggregative levels while guarantee simulation accuracyhas become the key research contents of network topology aggregative technique.The network topology aggregative algorithm proposed in this thesis includesredundant topology pruning, tree shrink and area shrink. It can determine thetopology aggregative level according to application demands to achieve topologyaggregative with different topology aggregative levels. Network topology ispreprocessed by the redundant topology pruning and tree shrink. Redundanttopology pruning aims to cut off the redundant nodes and redundant links fromnetwork topology. Tree shrink aims to shrink the internal nodes of tree areas to theroot nodes and route the internal packets of tree areas properly. Based on nodeweights and topology connective features, area shrink aggregates the areas whichmeet the aggregative conditions into a node or a link. Moreover, this thesisproposed an evaluation method of network topology aggregative algorithm usingAHP. This evaluation method can integrally evaluate network topology aggregative algorithm by virtue of taking the topology aggregative level and the simulationaccuracy into account and combining with the time complexities of topologyaggregative algorithm.To demonstrate the performance of the proposed algorithm, we organizedexperimental analyses using the network topological data provided by the VenusCompany in2009. The results show that the proposed network topologyaggregative algorithm gets better balance between simulation accuracy andtopology aggregative level which has higher topology aggregative level in thecondition of higher simulation accuracy. It achieves the best optimal performancewhich is about0.868when topology aggregative level is about0.69.