Abstract
The Adaptive Stochastic Rate Allocation (ASRA) algorithm for ABR traffic management in ATM networks is presented. As the dynamics and parameters of the traffic flow process are poorly known, the adaptive feedback gain depends not only on the parameters' estimates, but is also modulated by current uncertainty of the parameters. Therefore, the control algorithm expresses its dual nature by simultaneously performing two functions: ceaseless estimation of parameters and the process control. The allowed cell rates of the ABR sources are dynamically shaped by the feedback messages based on the instantaneous queue size in the ATM switch output buffer. The ASRA scheme is evaluated via simulation testing in comparison with the ERICA and BECN schemes. The requirements tested included: utilization, queuing delay, queuing delay variance, and queue size. The results of a simulation study suggest that the ASRA scheme can provide for higher priority VBR traffic, while retaining almost the same average utilization, considerably shorter average queue length, average queuing delay and particularly queuing delay variance in the ATM switch output buffer. The ASRA scheme's fairness in max‐min sense along with fast transient response is demonstrated.