For software development, especially massive software systems, a waterfall process is used traditionally. A waterfall process can be highly effective on the condition that a master plan is fixed and the possibility of changes and uncertain rework is low. However, in software development projects, many kinds of reworks occur corresponding to uncertain requirement changes and program bugs. In addition, with the advent of cloud-based software platforms and continuous development operations, it is possible to develop a software system while operating the system. To respond to this situation, software development projects often adopt an agile process. Agility may allow conditional response to uncertain rework, yet at the same time it may be difficult to control the achievement of known project targets. Recently, many cases of adopting mixed processes including waterfall and agile have been reported in the massive software development projects. In this paper, we argue that the mixed process architecture should be designed, considering the scale of the targeted software project, the culture of organization, the probability of uncertain requirement changes, and so on. This paper proposes a methodology of evaluating the impact of waterfall, agile, and mixed project architectures by using process simulation. A project architectural approach is evaluated with a simulator which includes a software reliability growth model and uncertain rework driven by requirement change and error propagation. The proposed methodology was applied to a development project for a simple shopping website. The results showed that the proposed methodology allows exploration of partial agile adoption depending on the nature of the system development project, including its scale and chances of change. For example, in this paper, if the scale of the project is small, the positive effect of increasing agility by adopting agile processes is low. On the other hand, if the scale of the project is large, the effect of increasing agility by adopting agile process can increase. Furthermore, it became clear that it is important to not apply an agile process blindly, but instead to design a mixed project architecture considering the number of errors and development schedule targets across the project scope.
The research on software project schedule prediction and controlling
