Recursive copy generation for multicast ATM switching
IEEE/ACM Transactions on Networking (TON)
Performance Analysis of a Multicast Switch Based on Multistage Interconnection Networks
INFOCOM '97 Proceedings of the INFOCOM '97. Sixteenth Annual Joint Conference of the IEEE Computer and Communications Societies. Driving the Information Revolution
Multicast ATM Switches Using Buffered MIN Structure: A Performance Study
INFOCOM '97 Proceedings of the INFOCOM '97. Sixteenth Annual Joint Conference of the IEEE Computer and Communications Societies. Driving the Information Revolution
Performance and Complexity of Multicast Cross-Path ATM Switches
INFOCOM '97 Proceedings of the INFOCOM '97. Sixteenth Annual Joint Conference of the IEEE Computer and Communications Societies. Driving the Information Revolution
On a Class of Multistage Interconnection Networks
IEEE Transactions on Computers
Nonblocking copy networks for multicast packet switching
IEEE Journal on Selected Areas in Communications
Hi-index | 0.24 |
This article explores the problem of efficiently multicasting packets on the baseline network in accordance with a given set of multicast communications. A baseline network is a multistage interconnection network (MIN) with N inputs and N outputs as well as log"2N stages of 2x2 switches. The baseline network construction with wraparound connections leads us to propose a new scheme for nonblocking packet multicasts. Previous approaches use a cascade of various MINs to multicast packets in a finite number of steps. Some others use single MINs to recycle and copy packets repeatedly until the multicast is done. Our proposed scheme exhibits a different approach from the previous ones, yet possessing many desirable features of theirs. Our scheme employs a log"2N-stage baseline network with wraparound connections. It is capable of accomplishing any multiple multicast in four finite passes. The first two passes replicate individual packets simultaneously. Then the subsequent two passes route these packets to the destinations. We demonstrate that paths created in the four passes are all link-disjoint so that nonblocking multicast is achieved. We further look into a few possible pipeline implementations of the proposed scheme. By unrolling some passes, we obtain a number of pipeline variations from the basic log"2N-stage baseline network. These variations allow us to process successive multicast runs in an overlapped fashion.