High speed bridges for high speed local area networks: packets per second vs. bits per second

Abstract

Progress in optical component design allows ever increasing channel capacity for Local Area Networks (LANs). Now that we have become familiar with data rates in the range of 10 Mbls the next step is already in front of us. Upcoming standards for High Speed Local Area Networks (HSLANs), prototypes and even products commercially available make HSLANs operating at data rates exceeding 100 Mbls move into the focus of interest. During the last few years many problems with respect to the design of MAC-protocols for these communication systems have been solved. It turned out that considerable modifications of protocols designed for LANs could not be avoided. However, there are still many problems to solve with respect to high speed bridges interconnecting LANs and HSLANs. In this paper we compare the performance of bridges to the pe?formance of the H S O W itsev. The results show that state-of-the-art bridges are very likely to become the bottlenecks in future high speed backbone networks.

Interconnection of high speed token ring LANs: a switch-connection approach

“…Today's state-of-the-art bridges are not able to forward more than a few thousand packets per second [4]. If we consider the interconnection of 100 Mb/s FDDI token ring LANs with packet length of 1,000 bits, the bridge processing time can be as large as 50.…”

“…The bridge processing time includes the frame forwarding, learning of the station addresses, and resolving possible loops in the topology by participation in the spanning tree algorithm [3]. The bridges that are commercially available today generally are not able to process packets as fast as they can be forwarded by high speed LANs [4]. An alternative to developing much faster bridges for preventing them from becoming system bottlenecks is to use non-store-and-forward interconnecting devices.…”