Towards a Global Software Development Distribution Model: Empirically-based Model Building for Distributed Software Development Master thesis

In recent years, global software development has become a common practice. Technological advances and expected benefits such as saving personnel costs and accessing a larger pool of resources have led to software development that is now often spread across several locations all over the world. However, distributed development also entails a number of risks and problems, for instance in terms of communication and knowledge transfer between sites, resulting in project failures. The challenge in global software development thus consists of maximizing the expected benefits while minimizing its risks and problems. A crucial factor for this is work allocation, i.e., the assignment of tasks in a software development project to the available sites. The goal of this thesis is the analysis of work allocation practices and criteria for the task assignment decision. The results of this analysis will then form the basis of an initial decision model that can support work allocation by providing a weighted list of assignment alternatives. The content of this thesis consists of three parts: First, a literature study analyzes experience reports, empirical studies, and other literature in order to identify factors that influence the work allocation decision. Second, a qualitative interview study with practitioners identifies the state of the practice in distributed development and refines the results of the literature study. Finally, a generic decision support model for distributed work allocation is developed and implemented. Based on the results of the literature and interview studies, the generic model is then instantiated into a specific decision support model that can be used to make suggestions for task assignment. The interview study resulted in a set of sometimes unexpected findings. For instance, there seems to be a difference in work allocation between software service companies (with mainly individual development projects for customers) and companies that develop their own software products: In product companies, task assignment is mostly based on expertise and on specialized teams, whereas in software service companies, the focus seems to be more on the availability of resources. Another surprising result was the fact that even though for most of the companies, cost was one of the main criteria for initiating global software development, the majority did not consider the cost rate per site as an important criterion for work allocation. The main contributions of this thesis are an analysis of the state of the practice in distributed work allocation, identification of a list of factors that influence task assignment, and a generic model – including a prototype implementation – for decision support in distributed work allocation.