Increasing multimedia applications such as teleconferencing and video-on-demand require the Internet to effectively provide high-performance multicast support. One of the promising solutions for multicasting is called end-system multicasting, in which the multicasting is conducted either at the end systems or at a dedicated Multicast Service Node (MSN). A MSN is responsible for a number of end systems. For performance purposes, MSN is a promising approach. Therefore, it is critical to design the architecture of MSN in general, and the queuing and scheduling on MSN in particular, such that it provides efficient multicast support. Previous work on multicast queuing and scheduling have implicitly assumed that multicasting is done at the core routers, a situation which is unlikely to happen in practice. In this paper, we designed a high-performance queuing scheme for MSNs to support both unicast and multicast and to achieve high link utilization and high switch throughput with a reasonable cost. The main approach is to use per-vector queueing on-demand at the ingress of the MSN. To verify the above design, simulation results were provided and analyzed. The results from this paper can be used as a part of the guidelines for building next generation multicast switches.
展开▼
