摘要
本文详细分析了随机早期检测RED拥塞控制机制算法的思想,阐述了RED的不足,提出了一种改进的动态自适应算法D-RED。该算法设置一个平均队列大小的目标值,在实时平均队列大小小于目标值和大于目标值时使最大丢包概率动态地减去和加上一个通过数学方法推导出的同路由器上一次丢包概率有关的值,以达到更好地控制网络拥塞的目的。利用NS2设计多组仿真实验对RED和D-RED性能进行比较,通过对吞吐量、延迟、振荡、丢包率等多项性能指标分析和比较表明,D-RED算法相对于RED算法使网络丢包率降低了15.7%,网络吞吐量提高了9.44%,而且能够在一定程度上保护TCP数据流,使其得到公平的网络资源。
The paper first analyses the design principle and algorithm implementation of the RED congestion control mechanism in touters, describes the defects of RED, and presents an improved adaptive algorithm D-RED. It sets a target value of average queue sizes, makes the maximal drop probability dynamically minus or plus a value, which is educed by mathematical methods, correlative to the former drop probability when the instantaneous average queue length is less or larger than the value of the target average queue. Then the paper designs many experiments to compare the performances of the RED and D-RED algorithms based on NS2. Finally, the paper analyses and compares the simulation results about the performance standards such as throughput, delay, variation, packets drop ratio, etc. It is concluded that D-RED can reduce the drop probability by 15. 7% and the throughput by 9.4,1% compared with RED. To some degree, TCP streams are protected so that they share fair network resources under D-RED,
出处
《计算机工程与科学》
CSCD
2007年第5期45-49,共5页
Computer Engineering & Science
