Over the last decade, emerging information communication technologies have changed our stereotype of manufacturing and service companies. Now products equipped with embedded systems can be wirelessly networked, which leads to gathering and analyzing product status, and taking appropriate actions for maintenance operations during product lifecycle in an ubiquitous way. In this environment, it is necessary to determine the appropriate memory size of embedded systems for minimizing total maintenance system costs because the memory cost is a main cost factor for implementing the ubiquitous maintenance environment. We call it memory size decision problem in this study. We have formulated this problem with a non-linear model having constraints. The decision variable is the memory size of each embedded system. To solve this problem, we have proposed a meta heuristic search method based on genetic algorithms. To show the usefulness of the proposed heuristic, we have carried out computational experiments. (C) 2008 Elsevier Ltd. All rights reserved.