What should be the goal for ATM

Two key features of asynchronous transfer mode (ATM) statistical multiplexing and continuous bit rate are analyzed. Both are claimed to provide the economic edge for ATM over other transmission and switching paradigms. It is argued that they have little chance to succeed in a high-speed environment, and even if they can be implemented, their benefits are hardly worth the complexity they add to the network. The trade-off of cell-granularity efficiency, in terms of bandwidth allocation, utilization, and billing, versus acceptable implementation complexity is discussed from the point of view of traffic management. A quantified-bandwidth channel and peak rate subscription are suggested as new goals for ATM.<>