Component selection strategies based on system requirements' dependencies on component attributes

Abstract

Rapid increases in systems complexity have raised the need to exploit the "design & reuse" principle to its full potential. The proposed research is targeted towards component reuse, specifically towards component selection. We assume a component specification method has been chosen and a component library has been designed and built. The problem we address in this paper is choosing a subset of components from a library of components, such that the resulting integrated system satisfies certain requirements. Our proposed approach contains two main stages. First, we address those requirements that can help us reduce our search space and secondly, we perform an intelligent search in our reduced search space. In the second stage we apply a Greedy approach for selecting components from our reduced search space. The challenge here is assessing how well a certain component satisfies the performance requirements of the target system, as these performance requirements usually refer to the system as a whole and not to individual components. To address this challenge we focused on mapping system performance requirements onto component characteristics. We will illustrate our proposed approach for component selection with a simplified example of selecting the components for a 4x4 mesh-based NOC (Network-on-Chip) architecture.