ATM Concepts, Architectures, and Protocols

Ronald J. Vetter, in CACM 38(2) Feb. 1995

Summary | Relevance | Rating | Back to Index

Content Summary

ATM was designed to be cost effective, scalable, compatible with existing networks at all scales, and easily implement multicast. Virtual-channel and virtual-path mechanisms are used for routing and to multiplex multiple paths into a single logical channel if desired; the VPI and VCI routing protocols are lightweight since they must be handled in hardware, and due to the high bandwidth of optical media the switch latency is now in the critical path. The values chosen for various parameters (bandwidths, cell sizes, etc.) often represented compromises between the US and European groups, both of whom wanted to retain compatibility with existing infrastructures.

The ATM reference model contains the adaptataion layer (AAL), which provides network models (connection-oriented, connectionless, time priorities or not); the ATM layer, which mostly does routing and control signalling; and the physical layer, which defines signal levels on various transmission media, etc. The two most widely accepted physical standards are currently SONET and HDS.

Remaining challenges for ATM include good emulation of existing LAN standards (e.g. IEEE 802.x), hardware incompatibilities among vendors, QOS scheduling in software (ATM provides the mechanism for doing this but no policy), creating fast switches, and implementing successful flow and congestion control.

Relevance to Multimedia

Lest ATM and optical networks be viewed as a magical cure-all for bandwidth- limited applications or software requiring hard QOS guarantees, this article highlights some of the challenges and difficulties that will accompany the transition, and presents some guidance to the maze of standards and protocols that the term "ATM" currently comprises.

Rating

4 out of 5: lots of useful tables and a good discussion of the real-life challenges faced by ATM.

Armando Fox (fox@cs.berkeley.edu)