Improving socio-technical coordination in development teams to avoid software failures

点击翻译按钮获取中文摘要

Software development is inherently a collaborative endeavour. Large teams of developers need to effectively coordinate their efforts in order to successfully manage frequent changes in the code, to integrate their work and to avoid failures. However, coordinating the contributions from large teams, often of hundreds of people who may span multiple offices, cities, or even continents, is very challenging. The goal of this research is to improve developer collaboration by actively determining and presenting in the developers' development environment information about relevant collaborators so that failures caused by changes to code can be avoided. In our methodology we conduct field studies of developers, and develop a model of and associated measures of socio-technical coordination, as well as techniques that use project historical data to determine relevant collaborators in the project. This research is conducted in collaboration with our industrial partner, the IBM RTC development team (also known as the Jazz team). We analyze software failures that can be traced to changes in the code, and the coordination patterns of developers involved in those changes. Based on measures for socio-technical coordination developed in this project, we develop techniques for the identification of coordination patterns that significantly relate to failures. We expect to implement our techniques into a collaboration recommender prototype that automatically identifies those developers involved in coordination patterns that significantly relate to failure-inducing changes. We will evaluate our techniques and developed prototype through user studies with the IBM RTC team. This research will bring contributions to both research and practice. For researchers, we will develop a means of describing and measuring socio-technical coordination as related to software failures, as well as techniques that can be leveraged to automatically identify collaborators that are critical to avoid software failures. By incorporating our prototype into the Jazz project, the Jazz team can use it in their daily collaboration activities to enhance its existing support to distributed collaboration of software teams, and to minimize the time handling coordination breakdowns.