Abstract

Considering a computational service as a set of processes, several issues can impact its performance, such as failures and shutdowns. Many strategies can be used to reduce this impact as the assignment of one service in several machines or the distribution of machines in different locations by respecting some constraints such as process dependency and capacity. The Machine Reassignment Problem is a hard problem that consists of a set of machines with associated resources and processes already assigned to these machines. The goal is to obtain a redistribution of the processes according to some optimization criteria, satisfying a set of constraints. In this work, we propose an efficient collaborative local search algorithm to solve the Machine Reassignment Problem. We pay special attention to designing an easily understandable algorithm that requires less computational resources than other more sophisticated well-known approaches in the literature. We show that our approach is effective using the ROADEF competition instances as a benchmark and that can obtain high-quality solutions.