A wide range of applications such as HDTV, VOD, and teleconferencing in today's networks require multicast transmissions. So, it is needed that ATM switches have the ability to handle multicast traffic efficiently in order to decrease the amount of traffic in networks. It is well known that head-of-line blocking limits the maximum throughput of an input queued switch to 58% for unicast traffic. Similar throughput limitation arises for multicast traffic.
In this thesis, we propose a novel multicast cell scheduling algorithm, called CFTR(Conflict-Free Time Reservation), in input-queued ATM switches. While the existing multicast cell scheduling algorithms have low throughput because of operating in FIFO fashion, CFTR increases the throughput by using non-FIFO input queues and the advance reservation method for contention resolution, for which each input/output keeps a reservation table representing whether time slots in the future are reserved or not. CFTR can operate at high speed because of independent output scheduling and a few iterations of reservation processes in each time slot.
The performance of CFTR is compared with existing unicast and multicast algorithms by simulation. Simulation results show that CFTR performs better than other unicast and multicast algorithms even with the small size of reservation tables and a few iterations of reservation processes.
