A discrete simulation model for assessing software project scheduling policies

Good project scheduling is an essential, but extremely hard task in software management practice. In a software project, the time needed to complete some development activity is difficult to estimate. Often, the completion of activities is delayed due to unanticipated rework which is caused by feedback in the process. In this paper, we show how process simulation can be used to support managers in finding good schedules for their software projects. We present a novel, stochastic simulation model which is tailored to the special dynamics of software projects, and which explicitly takes a scheduling strategy as input. The model represents task assignments, staff skill levels, component coupling, and rework caused by design changes. The simulation model is implemented in the ModL language of the generalpurpose graphical simulation tool EXTEND. As an illustration of our simulation model, we study the performance of various list policies for a small sample project. The simulations quickly show the impact that the choice of the list policy will have on the progress and completion time of the sample project. To explain the performance difference between the list policies, we use the simulation traces to provide a detailed analysis of the task assignments which actually occur in the simulations.