高级检索

    高速路由器中一种有效的组播交换排队机制

    An Efficient Queuing and Switching Scheme for Multicast in High-Speed Routers

    • 摘要: 许多Internet新业务产生的网络流量都属于组播类型.高速路由器中组播交换队列设计是解 决组播数据在复杂网络中有效传输的关键问题之一.理想情况下,为保证服务的公平性及有 效性,不应该在排队和调度时对组播和单播人为地进行区分.在二者共存的情况下,如何进 行公平地排队、交换是亟待解决的问题.基于共享存储交换结构,提出了一种有效的组播交 换队列设计方案,同时提供了相应的与组播比例和平均扇出相关的缓存管理机制和队列调度 算法.仿真结果表明,与在Cisco高端路由器中广泛应用的ESLIP设计方案相比,在输入负载 大于80%的重负载环境中,该系统能够获得更低的平均时延和更小的系统丢包率.

       

      Abstract: Many emerging Internet applications are multicast traffic. The queuing and switc hing scheme for multicast in high-speed routers is one of the key elements to en able the multicast traffic traverse the complex network environments. Ideally in order to guarantee the fairness between the unicast and multicast traffic we sh ould not differentiate them deliberately when queuing and scheduling. Since the increase of multicast traffic will exhaust the resource, the service for unicast traffic will be influenced. When the two types of traffic coexist, how to provi de the fairness between them from the point of queuing view desiderates to be re solved. Based on the shared memory switching architecture, an efficient queuing scheme for multicast packet switching routers is proposed. And the corresponding buffer management and scheduling algorithm according to multicast ratio and fan out are also presented based on the proposed queuing scheme. On three phases, wh ich are the input buffer, the shared memory and the buffer of egress NP, the ban dwidth and buffer allocation for multicast packets is decided according to the m ulticast traffic ratio and the mean fanout. Performance evaluation of the propos ed system shows that better average delay and loss performance can be achieved i n comparison with the ESLIP system under heavy offered load, which is applied fo r high-end routers of Cisco.

       

    /

    返回文章
    返回