Normalized Time and its Use in Architectural Design.

摘要

We began with the idea of normalized analysis: that the cost of a modification is just as important as its benefits. We have extended the model of normalized time to multiple groups of modifications. We then analyzed the results of the simplest, greedy, algorithm as a tool for selecting the best set of modifications. In doing so, we find that the greedy algorithm is provably a suboptimal algorithm, even for the very simple types of modifications considered here. Nevertheless, since it is a conservative algorithm, it is still useful as a starting point for further selection. By running the fast and simple greedy algorithm, we can select many of the same modifications that would be found by any better algorithm, thus reducing the number of choices that the other algorithms must make. With this theoretical basis, and the results of an initial algorithm, it may now be possible for computer designers to select, in a more analytical manner, which of the multitude of potential modifications to include in a computer system. (kr)