Measurements for managing software reliability

Abstract

Accurate software reliability evaluation requires the collection of comprehensive and consistent data sets on the ongoing software project. We define some measurements that can be performed on the software during its development or in operation to help evaluating and managing its reliability, and discuss organizational and feedback aspects. Since failure data are difficult to collect, the raw data set may include extraneous data: validation is thus needed before processing. Emphasis is put on near-term objectives, implying timely and efficient feedback for the ongoing project. However, as reliability measurement is to be considered in the first step in a software reliability improvement program, we report some success stories in which improvement programs have increased productivity and reliability at no extra cost or even with cost reduction.