Knowledge Based Optimization for Distributed Real-Time Systems

The design and the implementation of distributedreal-time systems has always been a challenging task. A centralquestion being how to efficiently coordinate parallel activities bymeans of point-to-point communication so as to keep global consistencywhile meeting timing constraints. In the domain of safetycritical applications, system predictability allows to pre-computeoptimal scheduling policies. In this paper, we consider a largerclass of systems represented as compositions of timed automatasubject to multiparty interactions, for which an implementationmethod for distributed platforms and based on intermediatemodel transformation already exists. To improve this approach,we developed specific static analysis techniques that, combinedwith local and global knowledge of the system, checks particularconditions that enables to decrease the number of messagesexchanged in the system for executing each interaction, as wellas to remove unnecessary scheduling overhead in some cases