Scheduling arbitrary number of malleable tasks on multiprocessor systems

Abstract

Abstract The problem of scheduling n tasks in a multiprocessor system with m processors to minimize the makespan is studied. Tasks are malleable, which means that a task can be executed by several processors at a time, its processing speed depends on the number of allocated processors, and a set of processors allocated to the same task can change over time. The processing speed of a task is a strictly increasing function of the number of processors allocated to this task. The earlier studies considered the case n ≤ m. This paper presents results for arbitrary n and m including characterizations of a feasible domain and an optimal solution, polynomial time algorithms for strictly increasing convex and concave processing speed functions, and a combinatorial exponential algorithm for arbitrary strictly increasing functions.