The design, implementation and initial evaluation of an advanced knowledge-based process scheduler

Abstract

Traditional operating systems control the execution of programs regardless of how often they are run. This raises the question: can't the often run or the often used programs provide better performance if an operating system had an ability to optimize their execution behavior based on a knowledge the operating system had obtained from their previous execution(s)? In this paper, we integrate this ability into a part of an operating system called a process scheduler and examine its cost and benefit. Our initial evaluations show that the cost involved in our scheduler is small and the processing time can be reduced by using this scheduler.