Effectiveness of the FDDI-M protocol in supporting synchronous traffic

Abstract

Timed token networks such as fiber distributed data interface (FDDI) networks have been widely deployed to support synchronous traffic. However the medium access control (MAC) protocol of FDDI allows transmission of synchronous messages up to at most one-half of the total bandwidth of the network. Shin and Zheng have proposed a modification to the FDDI MAC protocol, called FDDI-M, which can double a ring's ability in supporting synchronous traffic (K.G. Shin, G. Zheng, IEEE Transactions on Parallel and Distributed Systems 6 (1995) 1125–1131). It is widely known that the ability of timed token protocols such as FDDI to guarantee synchronous message deadlines is very dependent on the synchronous bandwidth allocation (SBA) schemes used, but the original paper does not address this issue. In this paper, we will compare the ability of FDDI-M to support synchronous traffic under different SBA schemes with that of FDDI. We use a new taxonomy of SBA schemes based on the strategy used to partition the synchronous bandwidth, and present an analytical study of the timing properties of the FDDI-M protocol using the worst case achievable utilization (WCAU) as the performance metric. The results show that while FDDI-M improves the WCAU values under one class of SBA schemes, its performance under the other category of SBA schemes is mixed. We also perform extensive simulation to study performance of FDDI-M for MPEG video traffic, and conclude FDDI-M does outperform FDDI significantly at heavy load. The effect of SBA schemes under overload conditions is also shown to be relatively minor, with the local SBA schemes actually performing better than the global schemes.