Tuning parameters to improve a heuristic method: more and better solutions to an NP-hard real-time problem

Abstract

The paper presents a tuned heuristic method to solve the NP-hard problem of assigning a set of preemptible tasks with placement, resource, communication, time and precedence constraints to a set of heterogeneous processors. Depending on the tuning parameters, two versions are described. They produce more and better (e.g. with a less loaded communication network) solutions than the untuned version. Time to obtain the first solution and the number of solutions found for different known problems are given.