Measuring the Progress of a System Development

Introduction

Nowadays software systems are being built using various technologies, appropriate to their application. Project leaders can choose between flexible agile methods (Martin et al., 2001), a spiral methodology (Boehm, 1988) for large and expensive projects, a waterfall model (Royce, 1970) for iterative developments, or prototyping originating from industry, among many others. Moreover, more rigorous approaches, like formal methods (Ladkin & Thomas, 2009), can be used for large, complex and critical system developments, e.g. those intended for space, aircraft or military sectors. Achieving high quality software is an aspiration of every development (Hayes & Jarvis, 1999). Hence, it is essential for software measurement to be an elementary part of the quality assurance program within every development process, including rigorous ones. When the measurements encompass several metrics, like size, complexity or effort, they can be regarded as reasonable quality indicators.

Since software changes are inevitable, it is crucial to be able to handle them efficiently (DeMarco, 1986). Monitoring the progress of the development benefits in better control and quicker reaction to the changes being made, e.g. by modifying the current modelling approach or refining the model. Given that some measures, like complexity (McCabe & Butler, 1989), can be computed early in the development cycle, they add significant value to managing the software process, as well as the software itself. Moreover, they can contribute to cost reduction and increase the maintainability and modularity.

We believe that the overall quality of the development process, including the systems fundamentals in the design phase strongly influences the quality of the final system. We are interested in modelling large and complex critical systems, embedded systems in particular, and focus on the initial stages of the development. We think that performing the changes at this point is efficient in the sense of time and money. Managing the complexity at this point influences the overall quality of the system and its enhancements. It also impacts specific quality attributes, such as dependability or, more specifically, maintainability or resilience. Therefore, it is possible to diminish the threat against dependability, for example, by constantly controlling the development quality with measurements.

The measurement activities in a software development project should already be applied by the designers. These are the personnel responsible for the modelling decisions, who have to take into account the specificity of the project, knowledge about the application domain and the environmental settings. Also at this point the analyst or quality assurance managers should have access to the collected data in order to present their findings to the project manager, who then makes higher level decisions about the project at its early stage. The testers, as a separate development group, should be aware of the outcome of measurements. The measurement findings can then be juxtaposed with testing results and, as a consequence, testers can propose possible changes.

The combination of well known methodologies that have been proven to be efficient and successful needs to be supported with a good engineering practice. In our research we benefit from formal methods and the Unified Modelling Language, which are merged with measurements for the purpose of quality assurance. A certain quality, including dependability, can be guaranteed and achieved by continuously monitoring the system with measures. The value of the system being created lies in system planning, anticipation of future needs and feasibility of redesign. However, this is not to be accomplished without the innovative use of technologies and continuous supervision of development progress.

One of our goals in this chapter is to show how generic refinement patterns are reflected in measurements. Moreover we want to demonstrate how design decisions influence the measurements and the quality of the system being constructed. Specifically we concentrate on evaluation of progress diagrams and pattern-based development with respect to size and complexity measures. The approach can be extended to tackle a more generic problem of statechart diagrams assessment. Our research results are targeting developers (particularly designers and system architects), as well as project and system managers. The methodology is intended for the design stage of the development and can assist in the system documentation.