There is suggestive evidence that the software industry has not yet matured as regards management of non-functional requirements (NFRs). Consequently the cost to achieve required quality is unnecessarily high.
The early research from our own group has been devoted to requirements management issues, such as prioritising and release planning. The current research performed by Andreas Borg has been documenting the state-of-practice of NFR and profiling of requirements engineering tools. The most recent study in development practices and process improvement was presented at the RE'06 conference in Minneapolis and shows how Ericsson is managing requirements, testing and design of capacity. In Ericsson's case capacity is the number of simultaneous mobile phone users in a network. The future goal of Andreas Borg is to describe a method for developing software-intensive systems that meets stated capacity requirements. Our non-functional requirements for a method working with capacity requirements are:
- It can manage with large systems.
- It can handle iterative development where a new release of the system is built on the current release to a large extent.
- It is model-based.
- It is empirically validated.
We are also co-operating with Åsa Hedenskog's research project on how usability requirements are handled in Ericsson's large organisation. In her coming PhD thesis Åsa Hedenskog describes the requirements engineering process as a communication between four levels of a Joint Control System.
Software quality in UML
Under the auspices of St Anna IT Research Institue Mattias Eriksson collected structural metrics and fault data from a large number of state-charts. The data is right now analyzed together with representatives of the contractor.
Together with colleagues at Umeå University we have replicated an empirical evaluation of Tony Gorschek's RAM method with a large number of students. We are currently publishing and analyzing the rich material.
Page responsible: Kristian Sandahl
Last updated: 2007-12-04