RPS类流量调度与成形算法中worst-case公平性能分析

The Fairness Study of SRPS Traffic Scheduling Algorithms

RPS类流量调度算法由于具有理想的延时性能,在具有QoS要求的调度算法中该类算法被广泛应用。为了使该类算法同时具有较好的公平性,将该算法与流量成形机制相结合(SRPS),即可得到理想的Worst-case公平指数(WFI)。在现有文献中都认为SRPS算法具有与WF2Q相同WFI,本文在分析RPS算法的基础上,重新给出了SRPS worst-case表达式,指出SRPS算法具有比RPS算法理想的公平性能,但还没有达到与WF2Q相同的Worst-case公平指数。

As RPS(Rate-Proportional Server) provides a methodology to design scheduling algorithms with ideal latency, it is widely employed by many traffic scheduling algorithms based on QoS applications. When combined with a shaping mechanism(SRPS), RPS can offer a batter worst-case fairness index(WFI) than RPS only. But until now, it is believed that the worst-case fairness index of SRPS was the same as that of WF2Q. In this paper, a revised SRPS worst-case expression is proposed. It is shown that SRPS is better than RPS but is not good as WF2Q in terms of worst-case fairness index.