Randomization Effect Measurement on the Fast Power Consumption Scheduler

Abstract

Power consumption scheduling is a problem that needs intensive computing time, so it is necessary to solve with a heuristic method. Most heuristic schemes, when no prior guideline is available, begin with the random initial selection and then iteratively refine the feasible schedule. This paper measures the effect of randomization ratio in the initial selection to the peak load reduction to improve the fast power consumption scheduler. The experiment focuses on the performance metrics of the number of tasks, the operation length, and the slack distribution to analyze their effect on the selection of better randomization ratio. The measurement result obtained by the prototype implementation shows that when the operation length and the slack increases, that is, there are sufficient number of selectable options for a power consumption schedule, the randomization ratio around 0.3 can best reduce the peak load.