The priority token bank in a network of queues

This paper takes a known approach for scheduling and admission control in integrated services networks, the priority token bank (PTB), whose mechanism and performance have been studied in a single queue, and extends the algorithm to address the challenges of implementing it in a network of queues. An integrated services network must handle traffic streams with disparate arrival processes and performance objectives, and efficiently supporting these streams becomes more difficult when performance guarantees must be met across a number of switches and when a stream´s burstiness can dramatically increase from source to destination. Three variations on the basic mechanism are proposed to address this problem: PTB-separate classes, PTB-minimum interdeparture time, and PTB-weighted fair queuing. It is shown that the PTB scales well to a network of queues, and that, under the constraint that performance requirements for guaranteed traffic such as packet voice, HDTV, and packet video must be met, the mean delay experienced by other traffic is much better with the priority token bank mechanisms than with other proposed algorithms