A statistical model for system components selection

The proposed research is aimed at enhancing system design productivity by exploiting the principle of “design and reuse” to its full potential. Specifically, we present a statistical model for selecting from a component library the optimal components for a network-on-chip architecture such that to satisfy certain system performance requirements. Our model is based on regression analysis and Taguchi's optimization technique. The model estimates the relationship between system performance and component attributes, to help the architect in the component selection process. Having such a model in the system design phase will allow the architect not only to make informed decisions when selecting components but also to exchange components with similar characteristics to fine tune system performance.