This paper provides four contributions to the study of optimization techniques for component-based distributed realtime and embedded (DRE) systems. First, we describe key challenges of designing component-based DRE systems and identify key sources of overhead in a typical componentbased DRE system from the domain of shipboard computing. Second, we describe a class of optimization techniques applicable to the deployment of component-based DRE systems. Third, we describe the Physical Assembly Mapper (PAM), which is a model-driven optimization tool that implements these techniques to reduce footprint. Fourth, we evaluate the benefits of these optimization techniques empirically and analyze the results. Our results indicate that the deployment-time optimization techniques in PAM provides significant benefits, such as 45% improvement in footprint, when compared to conventional component middleware technologies.
Krishnakumar Balasubramanian, Douglas C. Schmidt