An Experiment Evaluating Architectures of Software Measurement Systems

Abstract

A software measurement process helps to collect data in order to evaluate a software product on a continuous basis. However, the process can take a considerable amount of time. Project managers are faced with a decision how many measurements they can get in a certain amount of time, i.e. performance of a software measurement tool. This thesis describes a controlled experiment in academia that was aimed at evaluating performance of a software measurement tool. The results show how the architecture of a measurement system affects its performance. Finally, we can use several architectural strategies described in the thesis to minimize decrease in performance.