An queue has Poisson arrivals, independent identically distributed service times, and one server. If service has mean and , its mean busy period isThis follows by decomposing the busy period into the first service and the busy-period descendants of arrivals during that service.
An M/D/1 queue has a Poisson process of arrivals, a deterministic service time , and one work-conserving server. With arrival rate and , its equilibrium idle probability is : the stationary completed work per unit time is . Its output generally need not be a Poisson process.
Let be the number left immediately after departure , and let be the arrivals during the next service. ThenConditional on service duration , is Poisson with parameter , so
For a stable first-come first-served M/G/1 queue, write and . The stationary waiting-time Laplace transform isWith finite second service moment, differentiating at zero gives the Pollaczek-Khinchine mean formula. These are the transform and mean forms of the same queueing identity.
For a stationary first come first served M/G/1 queue, take independent service times independent of the Poisson process of arrivals. Let , and . Then the mean queueing delay and mean queue population areTo see why the second moment appears, a completed service of length contributes triangular area under the residual service time curve. The renewal-reward theorem gives mean residual duration , counting zero while idle. Poisson arrivals see time averages and Little's law then give . Solving this identity proves the formula.[ref]
For the same mean , deterministic service has , while exponential service has , doubling the mean wait at fixed arrival rate. With minute and per minute, these waits are respectively and one minute. Stability alone does not guarantee finite mean delay: for positive and infinite second moment, the mean FCFS wait is infinite even when .
Articles by others on the same topic
There are currently no matching articles.