A Feedback Macro-Scheduler

A multi-class macro-scheduler is described in this paper. The scheduler periodically determines the number of jobs from each class that should be activated to minimize a weighted sum of the mean system residence time without saturating the system. The computation is based on the estimated system workload in the next interval. Thus it is adaptive to workload variation. The service provided to each class (specifically, the mean response time) may be adjusted by changing the weight associated with the job class.
The scheme is based on mathematical modelling. The solution is obtained through the use of queuing theory, operational analysis and optimization theory. Exponential smoothing technique is employed to reduce the error of estimating the value of the model parameters. Simulation results show the scheme to be both stable and robust. Performance improvement over some of the existing schemes (the 50%, L=S and the Knee criteria) is significant under some workloads. The overhead involved in its implementation is acceptable and the errors due to some of the assumptions used in the formulation and solution of the model are discussed.