Towards evolutionary and systematic process modelling using components

There is growing recognition that successful software systems evolve. Similarly, the processes that produce software must also evolve in order to free evolutionary system growth. Systems Dynamics modelling and simulation has been used to support process improvement strategies; however ad-hoc modelling methods may cause failures of understanding that lead to failures of these strategies. If we are to build better, evolvable software development processes with predictable behaviour and outcome, we need to be able to use modelling and simulation in a more systematic way. This thesis describes an evolutionary modelling method that uses quantitative simulation to ensure close correspondence between a Systems Dynamics model and the behaviour of a software development process. Secondly, through two experiments, we show how componentisation allows us to evolve process models in a dependable way, by breaking processes down into components that are well understood, with predictable behaviour. We suggest that we will be better placed to design evolvable, flexible processes that make good use of complex strategies like distribution, concurrency and feedback if we can develop re-usable process components, with well understood and predictable behaviour in the software development domain.