Effective bandwidths at multi-class queues (Q1176436): Difference between revisions

An M/G/1 queue is used as model for a buffered resource, the customers being bursts of \(J\) different types each with their own mean service time \(\mu_ j\) and arrival rate \(n_ j\nu_ j\) where \(n_ j\) is the number of sources of type \(j\). By using well-known properties of the M/G/1 queue, it is shown that linear constraints of the type \(\sum^ J_ 1\alpha_ jn_ j\leq K\) ensure that the utilization, resp. the mean workload, resp. a given tail probability for the buffer content, will stay below a specified level (here the effective bandwidth \(\alpha_ j\) of a source of class \(j\) is defined in a different way for each of the three problems; for example for the utilization, it is just the traffic intensity of a class \(j\) source).