Software Architecture Measurement—Experiences from a Multinational Company

In this paper, we present our 4-year experience of creating, evolving, and validating an automated software architecture measurement system within Huawei. This system is centered around a comprehensive scale called the Standard Architecture Index (SAI), which is composed of a number of measures, each reflecting a recurring architecture problem. Development teams use this as a guide to figure out how to achieve a better score by addressing the underlying problems. The measurement practice thus motivates desired behaviors and outcomes. In this paper, we present our experience of creating and validating SAI 1.0 and 2.0, which has been adopted as the enterprise-wide standard, and our directions towards SAI 3.0. We will describe how we got the development teams to accept and apply SAI through pilot studies, constantly adjusting the formula based on feedback, and correlating SAI scores with productivity measures. Our experience shows that it is critical to guide development teams to focus on the underlying problems behind each measure within SAI, rather than on the score itself. It is also critical to introduce state-of-the-art technologies to the development teams. In doing so they can leverage these technologies to pinpoint and quantify architecture problems so that better SAI scores can be achieved, along with better quality and productivity.