The majority of ATM switches that have been proposed only support unicast (point-to-point) connections. Those supporting multicast (point-to-multipoint) connections tend to perform poorly, with acceptable multicast performance only achievable using an excessive amount of hardware. Because of the growing importance of multicast traffic, there is the demand for multicast switch designs which offer both low hardware complexity and high performance. This research investigates a class of multicast ATM switches called recycling switches which can satisfy both requirements. Recycling switch performance is studied using a simulated network model. The major performance parameters measured are the loss rate, mean delay, and delay variance of cells crossing through the switch under uniform and bursty traffic patterns. The reason recycling is not more widely used in multicast switches is the perception that it can lead to some multicast cells receiving lower quality of service than others. This research demonstrates a new priority-based approach to designing recycling multicast ATM switches which addresses this problem while maintaining low complexity and excellent scalability.
University of Newcastle Research Higher Degree Thesis