Adaptive ATM Access Control

One problem is how to admit the maximal number of these disparate users while maintaining service quality. Simply allocating users bandwidth equal to their maximal (peak) transmission rate, while straightforward, would be uneconomical for bursty users who may have an average rate at a small fraction of their peak rate. Methods that can statistically multiplex many bursty users would provide significantly greater network utilization. For example, a source with many short bursts and 5% utilization could potentially be multiplexed with close to 20 times more users on the channel than with peak rate allocation while maintaining the same quality of service level. But in order to guarantee quality of service, any statistical multiplexing technique must be robust to a range of call statistics. With this motivation, we more fully describe the problem.