针对网络信息量的不断增大造成的网络拥塞已经成为制约网络性能的主要因素的问题,研究了现有路由算法的改进,以缓解可能出现的拥塞,并提出了一种拥塞预知路由算法(CPRA)。该算法通过周期性地检测队列缓冲区占用率(BOR)来判断链路是否有发生拥塞的趋势,当队列BOR达到一定阈值时,则认为链路有发生拥塞的趋势,并根据局部的拓扑结构与链路状态快速计算备用路由,当链路BOIl达到一定阈值时启用备用路由转发数据包。在网络模拟器NS3上的仿真结果表明,相对于开放最短路径优先(OSPF)算法,CPRA能够显著提高网络的性能:网络吞吐率提高约50%,端到端平均时延减少0.3s。
Aiming at the problem that the network congestion caused by the increasing network information now becomes the main factor restricting the performance of networks, the improvement of existing routing algorithms was conduc- ted to reduce the possible congestion, and a congestion perception routing algorithm (CPRA) was presented. This algorithm performed the periodic testing of queue buffer occupancy rate (BOR) to predict whether the link is going to be congested or not. When the queue buffer occupancy rate reached a certain threshold, the link was considered more likely to be congested. A fast calculation of alternate routing was performed according to the local topology and link state, and the alternate routing was used to forward packets when the queue buffer occupancy rate reached a certain threshold. The results of the extensive simulation on the simulator NS3 showed that, compared with open shortest path first (OSPF) algorithm, The CPRA greatly improved the network performance: raising the network throughput by about 50% and decreasing the average end-to-end delay by about 0.3s.