【作者】 金镝； 王晓川； 金士尧；

【Author】 Jin Di1+, Wang Xiao Chuan2, Jin Shi Yao3(NATIONAL UNIVERSITY OF DEFENSE TECHNOLOGY,410073 China)

【机构】 国防科技大学分布与并行处理国家重点实验室；

【摘要】 本文针对互联网拥塞控制,提出了一种新的基于模糊理论的拥塞控制算法(FLCCA)。依据队列的缓冲区占用率,以及满队列时间作为语言变量输入,建立拥塞控制模型,通过模糊推理计算数据包的丢弃率。并且采用全局性队列和局部性队列相结合的拥塞控制方法,实现队列调度的模糊控制。NS-2的仿真结果显示,该算法相比传统方法具有更好的性能和鲁棒性,保证了在连接服务方面的公平性,改进了路由器/交换机的拥塞控制性能。更多还原

【Abstract】 With the deferent information becomes larger and network’s structure becomesmore complex, network’s loads always overrun network’s capacity, and it increases thepossibility of the congestion. If we could not control the congestion effectively and make itback to the normal state when the congestion comes, it will result in the serious congestion,even collapse. Therefore, it is more important to avoid and control the network’s congestion.Firstly, this paper designs a congestion control model, and introduces the basisconceptions of the model. And then we discuss the queue management based on the fuzzylogic. Secondly, we establish a fuzzy system which has the two imports and one export.Buffer performance and full queue time are the imports of the system. Loss of the queue isthe export of the system. In order to reduce the complexity, we predigest the subjectionfunctions. The established subjection functions are not unique, if it can reflect theimpersonality reality, we could consider it right.According to the shortage of the artery congestion control algorithms and theindeterminacy of the state information of the network, we can combine the fuzzy logic andcongestion control. In order to perfect the fuzzy logic congestion control algorithms, a fuzzylogic congestion control algorithm (FLCCA) is proposed which based on the bufferperformance and full queue time. It scouts the state of the buffer performance and full queuetime, and then it judges the congestion state of the queue by fuzzy reasoning, so we coulddeal with the data flow which is real congested. According to the diverse service type, itprovides the Diff-Serv and ensures that diverse service type can get the corresponding servicequality. It combines the whole congestion control and the part congestion control.At the last of this paper, we use NS-2 stimulation to stimulate the FLCCA. NS is amellow network simulate environment. We extend the NS that the TD-FIFO algorithm andRED algorithm are implemented in the primary simulator and we implement the FLCCAalgorithm in the extended simulator. Moreover, we compare the performance of the FLCCAalgorithm to the other algorithms like TD-FIFO and RED, and the results are analyzed andestimated. The NS-2 stimulation results show that the algorithm of FLCCA behaves betterperformance than that of other AQM algorithms. It preserves the fairness of all connections.It really improves the routers’ ability to deal with the network congestion.更多还原