Abstract: To improve the performance of network simulation, a network simulation method based on topology abstraction is presented. In this method, the large-scale network topology is abstracted into a small-scale one. By analyzing from the simulation result of small-scale topology, the original simulation result of large-scale topology is obtained. It is proved theoretically that the method can keep the simulation fidelity of the instantaneous network behavior parameters, such as queue length of router, packet drop rate, packet forwarding delay, and TCP sending rate etc. Experimental results based on NS2 show that when the degree of topology abstract is 16, compared to the traditional network simulation, the method can decrease the simulation running time by over 98%, yet the simulation results of network behavior are of high fidelity.

Key words: network simulation, performance of simulation, topology abstraction, simulation fidelity

摘要： 为提升网络模拟性能，提出了基于拓扑抽象的网络模拟方法。该方法对大规模网络拓扑进行抽象，形成小规模网络拓扑；通过对该小规模拓扑的模拟结果进行分析，形成原始大规模拓扑的模拟结果。理论证明了该方法能保证路由器队列长度、丢包率、数据包转发时延、TCP流量速率等瞬时网络行为参数的模拟真实性。通过基于NS2的实验结果表明，当拓扑抽象度达到16时，相对于传统的网络模拟，该方法降低模拟运行时间98%以上，而网络行为模拟结果仍具有较高的真实性。

关键词: 网络模拟, 模拟性能, 拓扑抽象, 模拟真实性