高级检索
    李士宁, 官骏鸣, 覃 征. KIR:一种TCP拥塞避免阶段的公平性算法及在GEO卫星环境下的仿真[J]. 计算机研究与发展, 2006, 43(12): 2048-2055.
    引用本文: 李士宁, 官骏鸣, 覃 征. KIR:一种TCP拥塞避免阶段的公平性算法及在GEO卫星环境下的仿真[J]. 计算机研究与发展, 2006, 43(12): 2048-2055.
    Li Shining, Guan Junming, Qin Zheng. KIR:A New Algorithm to Improve the Fairness of TCP Congestion Avoidance[J]. Journal of Computer Research and Development, 2006, 43(12): 2048-2055.
    Citation: Li Shining, Guan Junming, Qin Zheng. KIR:A New Algorithm to Improve the Fairness of TCP Congestion Avoidance[J]. Journal of Computer Research and Development, 2006, 43(12): 2048-2055.

    KIR:一种TCP拥塞避免阶段的公平性算法及在GEO卫星环境下的仿真

    KIR:A New Algorithm to Improve the Fairness of TCP Congestion Avoidance

    • 摘要: 传统的TCP拥塞避免机制对长时延链接存在歧视,随着竞争流的增加,TCP共享瓶颈带宽的公平性和有效性降低.在研究CR,IBK,CANIT等现有算法的基础上,提出了一个新的均衡公平性算法KIR (K and additive increase ratio)来纠正对长时延链接的歧视,新算法的思想是在拥塞避免阶段分别对长时延链接与短时延链接的窗口增加方案做了平滑修改,同时经过数学推导,给出一个能够取得较好公平性和带宽利用率的K的表达式.通过一系列仿真实验比较了不同公平性机制的性能,分析了它们之间的不同特性,结果显示KIR算法不仅能够提高TCP公平性,而且取得了比较理想的链路吞吐率.针对地球同步轨道GEO卫星环境,在NewReno,Sack,Tcpw三种TCP拥塞控制机制中对KIR算法的有效性进行了仿真验证.

       

      Abstract: The traditional TCP congestion avoidance mechanism has strong bias against links with higher round-trip delays. As the competing TCP connection increase, the fairness and utilization of the sharing link degrades dramatically. The CR, IBK, CANIT examined firstly, and then a new fairness algorithm “K and additive increase ratio” (KIR) is proposed to correct the bias against these long connections. The new algorithm in which a new arithmetic formula “K” is used for the first time smoothly modifies the long and short round-trip delay link congestion avoidance algorithm. A series simulation is chosen and the different algorithm characteristic is analyzed. With these modifications, the simulation result show that the algorithm not only can improve TCP fairness, but can obtain good throughput performance as well. Finally, the effectiveness of KIR is proved by the simulation combined with NewReno, Sack and Tcpw under the GEO satellite environment.

       

    /

    返回文章
    返回