Deployment automation with BLITZ
Brian P. Dougherty, Jules White, Jaiganesh Balasubramanian, Chris Thompson, Douglas C. Schmidt · 2009
Minimizing the computing infrastructure (such as processors) in a distributed real-time embedded (DRE) system deployment helps reduce system size, weight, power consumption, and cost. To support software components and applications on the computing infrastructure, the hardware must provide enough processors to ensure that all applications can be scheduled without missing real-time deadlines. In addition to ensuring scheduling constraints, sufficient resources (such as memory) must be available to the software. It is hard to identify the best way(s) of deploying software components on hardware processors to minimize computing infrastructure and meet complex DRE constraints. This paper provides the following contributions to the study of the deployment of software components to hardware in DRE systems: (1) we present an algorithmic deployment technique that minimizes the required number of processors, while adhering to real-time scheduling, resource, and co-location constraints, (2) we show how this technique can be augmented with a harmonic period heuristic to further reduce the number of required processors, and (3) we present empirical data from applying three different deployment algorithms for processor minimization to a flight avionics DRE system.