الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 32 |  |  |
| | | from | 32 | | |
Abstract
A new high performance ATM switch, called Fast Multicast Parallel-Banyan Switch (FMPBS), has been introduced ibr handling multirate traftic and higher bit rate services required for ATM cell processing specially in case of bursty data and multicast function with efficient use of bandwidth. The switch architecture is based on several key principles: non-blocking, self-routing, output queuing for delay and throughput enhancement, levels of service priority required for maintaining QoS, high level of parallelism for achieving scalability, reliability apd higher speed, and multicast function. The switch is modeled analytically and simulated. The experiment results are similar to the analytical results. The switch performance is evaluated in terms of cell loss probability, speed, delay, throughput, utilization, and maximum number of transmitted multicast cells in a time slot. Then it is compared with similar space division designs based on different approaches: Tandem-Banyan, Cbs network and multicast tree. The switch has been found to be of less cell loss rate (CLR) and delay. In addition, it provides a higher bit rate services independently from the multicast load. The multicast load has a great effect on the CLR and the queuing delay. Efficient multicast and broadcast routing mechanisms are required in order to reduce the multicast load and the communication costs for applications that send the same data to multiple recipients across a wide-area network. Efficiency is measured in terms of storage, bandwidth consumption, and traffic concentration: data packets or membership information should be sent over links that lead to either receivers or senders. Two different delivery trees, the collection of nodes (routers) and links that a multicast or a broadcast packet traverses, have been constructed by two new routing algorithms. The first tree is a Single Spanning Tree Broadcast Routing (SSTBR) to deliver broadcast packets with minimal network cost and minimal number of traversed links. The second tree is a single Minimal Bandwidth Group-Shared Tree (MBGST) per group, to deliver a multicast packet. The MBGST is a constriction of shortest-path multicast tree (SPT) in order to minimize the number of traversed links and to reduce the fanout. The performance measures, namely, bandwidth consumption, maximum fanout, storage, tree cost, maximum packet delay, and reliability are considered. The simulation results indicate a 45% reduction of the multicast load independent from the network growth. This realizes a greater traffic concentration.