基于网络演算的显示拥塞通知性能分析

Network Calculus Based Performance Analysis on Explicit Congestion Notification

网络拥塞控制是提升网络性能的重要手段,其研究受到广泛关注。显示拥塞通知(Explicit Congestion Notification, ECN)机制作为一种有效的网络拥塞控制协议已经开始在网络中部署。现有的ECN研究工作大多集中在交换机阈值调优算法上,缺少相关的理论研究,因而无法对算法设置的交换机阈值的最优性进行有效评估。通过利用网络演算分析交换机阈值和网络性能的关系,针对交叉流与目标流同时进入交换机并且不同流具有不同优先级的场景,提出一种分析网络性能边界的算法。实验结果表明,在不同场景下分析得到的性能是仿真性能的有效边界,验证了分析算法的准确性。

Network congestion control is an important means to improve network performance,and its research has attracted widespread attention.Explicit Congestion Notification(ECN)mechanism has recently been deployed in networks as an effective network congestion control protocol.However,most existing ECN research efforts focus on switch threshold tuning algorithms,lacking relevant theoretical research,and thus are unable to effectively evaluate the optimality of the switch thresholds set by the algorithm.By using network calculus to analyze the relationship between switch thresholds and network performance,an algorithm is proposed to analyze the network performance boundary for scenarios where cross traffic and target traffic enter the switch simultaneously and different flows have different priorities.The experimental results show that the performance obtained by analysis is an effective boundary of simulation performance under different scenarios,verifying the correctness of the analysis algorithm.