Abstract Decision Engine for Task Scheduling in Linux Based Operating Systems

Schedulers have improved and optimized the performance of various multi-threaded applications over many years. Schedulers are conceived as a part of user space task or realized as an algorithmic implementation within the operating system scheduler. However, there is no generic design in place to enforce a scheduler design targeting a selected group of tasks in the operating system. In this paper, we present a novel approach to generalize an abstract decision engine, which would select the appropriate scheduler design based on the user and system constraints. This paper also provides a case study on IRS(Iterative Relaxed Scheduling) framework. The evaluation of this case study provides a foundation for the generic decision engine, which would alter the behavior of selected tasks in the operating system.