混合模型仿真大规模网络实现

在仿真的效率和可扩展性方面，单纯的数据包仿真已无法满足网络发展的需要．该文将流仿真模型嵌套到数据包仿真的程序中，实现了流仿真同数据包仿真的集成．在仿真中，流网络通过流链路链接到包级仿真的Ns节点上，这种采用集成混合模型进行的网络仿真，既保持了数据包仿真的参数精确性，又提高了仿真效率．实验结果表明：采用混合模型仿真TCP流可以获得正确的动态特性；混合模型仿真可以提高网络的仿真效率，提高的程度与采用数据包仿真流所占整个TCP流的比例有关．

Implementation of Hybrid Model Simulation in Large Networks

The fast growth of computer networks gives packet simulation heavy computation load, making it unsuitable for truly large scale models. In this paper, the integration of fluid network simulation with packet simulation is realized by embedding fluid model in the program of packet simulation. The fluid network is connected by fluid links to Ns nodes at the packet level of the simulation. The proposed integration, which is termed hybrid simulation, shows a good performance of fluid models while preserving detailed packet level information for selected packet traffic flows. Finally, a set of experiments are conducted to test the hybrid model. The results show that the hybrid model simulation can ensure the accurate transient behavior in simulating TCP flow. The hybrid model can improve network simulation effectively. The effect is related to the proportion between packet simulation flow and TCP flow.