分组公平排队算法的仿真研究

Research on Simulation of Packet Fair Queueing Algorithms

  • 摘要: 对ATM复接器和输出缓存ATM交换机采用各种分组公平排队算法情况下的信元时延进行了仿真。仿真结果表明,SCFQ对连接之间的隔离性比其他算法差;若一个CBR速率组内没有流量超过其预约带宽的连接,则用Per_Rate_Grouping方式实现WF2Q+和不用这种实现方式,从最大时延和平均时延来看,相差不大。提出了一个调度CBR、VBR、ABR和UBR的调度器结构,既能满足CBR对时延的严格要求,又能保证VBR、ABR和UBR的吞吐率,同时还具有较低的实现复杂性。

     

    Abstract: In this paper, the cell delay in ATM MUX and output-buffered ATM switch is simulated, Which adopts various packet fair queuing algorithms. The simulation results are analyzed, which show that the isolation property of SCFQ(self clocked fair queuing) is weaker than others. If a CBR group has no connection in which traffic exceeds reservation, as to the maximum cell delay and cell average delay, the difference between per_rate_grouping implementation of WF2Q+ and no_per_rate_grouping implementation of WF2Q+ is negligible. According to the results, a scheduling structure dealing with CBR. VBR. ABR and UBR is proposed, which satisfies the strict delay requirement of CBR and guarantees the throughput of VBR. ABR and UBR. The implementation complexity of the scheduler is also discussed.

     

/

返回文章
返回