Optimizing the cost of an heterogeneous distributed platform

Distributed platforms become heterogeneous in more and more domains, as heterogeneous computing (HC) onto grids or reconfigurable factories in the industry. For production grids and factories, it is mandatory to control and optimize the economic cost of a such platforms regarding performance objectives. We present in this paper a study which purpose is to optimize the size of such environments depending on the workflow to execute or product to realize. The target platforms are either micro-factories, sized to manufacture products at the micrometric scale, or the heterogeneous computing domain where the key point is to reserve processors of an execution platform onto a grid to compute workflows like medical imaging applications. Thanks to the sizing of the platform, optimal or not, scheduling a workflow in HC environment or a production in the micro-factory is easy because the size of the platform already takes the performance constraints into account. In this paper, we present general results on the platform size optimization. Numerical results are also presented to illustrate 3 cases of our study.