Abstract
In the nonpreemptive highest-response-ratio-next discipline, the priority of a job is defined as the ratio of the time it has spent in the system to its service requirement. We evaluate the performance of this discipline by deriving an expression for the average waiting time of a job conditioned on its service requirement for arbitrary service time distributions. This expression is then evaluated analytically for constant service times and by a combination of analytical and numerical methods for hypoexponential, exponential, and hyperexponential distributions. For the latter, our exact result is compared with Br inch Hansen's approximation. We plot our results together with the waiting times of the first-come first-serve and shortest-job-first disciplines on which most batch-processing schedulers are based. In comparison to these two schemes, the highest-response-ratio-next discipline appears as an attractive alternative since it assures fast service to short jobs without excessively delaying longer ones