Learning and Assessing Risks for Enhancing Efficient Stream Computing by Fitting Computations to Cores

Our overall goal is to create a method for creating efficient and portable stream computing solutions. The central question that we need to solve in this context is how to make quick and accurate decisions on where to map workload in the computing infrastructure. Optimally, the execution of the workload should be such that the intended target objective is reached e.g. maximal performance or minimal energy consumption. In order to measure the effectiveness of a mapping decisions, a new metric Fitness is defined. The Fitness metric is aimed to be a universal metric useful 1) at the software design phase, 2) for runtime scheduling, and 3) for the implementation of hardware itself.