Using the CMM to evaluate student SE projects

The Computer Science and Software Engineering degree at Swinburne has a strong emphasis on student projects. Students undertake a sequence of three such projects: a small team project in Ada in first year; a small team project with database orientation in second year; a large team project in Ada, with a real-time distributed system orientation in third year, after students have spent a year on placement in industry. Our experience has been that the first two projects have generally served their purpose quite well. The technical simplicity of the projects has not detracted from the process-oriented learning goals we have for our students, and the small team size (generally around four) has enabled management to proceed relatively smoothly. We are pleased to concur with the literature on the basic success of the small team project model. However the situation with the large final year projects has been rather different. To date (over three years) we have not had any project that we could happily describe as exemplar of our ideals. Often the student experience has been of desperate measures taken to ensure that software is delivered, quite against the principles of their software engineering education. Our desire to encourage the development of distributed real-time systems, integrating a variety of heterogenous commercial software products (such as Alsys Ada, Oracle, Telesoft TeleUse) has resulted in real-life frustration, as students have battled with technology rather than appreciated wise process. Further, we have not discovered an assessment scheme ideal for the purpose of the project. In summary, what promised to be an interesting learning experience for our students, aimed at consummating the thrust of the degree, has to some extent failed to deliver. We feel that we need to reconceptualise the large team project in our context. The paper describes an attempt to address the development of a framework for the conduct and assessment of large project subjects, inspired by the SEI Capability Maturity Model for Software. We observe that the primary requirements for projects concur well with the requirements for level 2 (Repeatable) in the CMM. We describe an action learning experiment, still in the embryonic stage, where we impose a CMM level 2 framework on the conduct of student projects. We describe an initial implementation, in the context of a coursework masters degree in Software Engineering.