The motivation for the use of ATM techniques in the local area stems from four cohesive points of view; ATM is:
.
The technological and performance arguments for networks based on a general topology interconnection of ATM switches have been, and continue to be, fully represented in the literature. Switches with 160 Mbit/s line rates and aggregate capacities of several Gbit/s are now feasible with current technology, offering a packet switched service at rates previously only associated with STM networks. Small scale switches for private and LAN use are now becoming available; in our environment we use switches varying in size from 4 by 4 to 16 by 16 developed within the Fairisle project [14]. Designs for larger switches such as required for public networks present interesting scaling problems.
However, our consideration of ATM based networks in general (rather than simply B-ISDN), also leads to the consideration of networks based on different cost, technology and management tradeoffs, which serve to extend the spectrum of solutions offered by ATM networks. Examples include DQDB extended to support virtual circuits, the ATM ring recently proposed as a CPN [13], and the 500 Mbit/s CBN [6] (a network in everyday use between our two laboratories). These ATM technologies encompass a variety of physical layer and cell header formats, but are able to inter-work at the ATM layer.
The flexibility of ATM comes from the fine grain of multiplexing present in the ATM layer, which allows a large delay-insensitive packet to be pre-empted by higher priority or time-sensitive traffic. While it is true that this allows ATM to provide an integrated mechanism to carry traditional STM and PTM services, the ATM multiplexing can also offer a flexible interfacing mechanism to end-systems and hence a much richer range of possible services can be provided to applications. It is this latter aspect which we have developed at Cambridge.