There is an increasing amount of non-traditional software being developed by developers, i.e. software developed by people not formally trained in software engineering practices and developed in environments that place different constraints on software development than are found in more traditional IT software development environments. Software developed for Computational Science and Engineering is a prominent and important type of non-traditional software. Computational science and engineering software supports the investigation of topics that impact our vital national interests (e.g. cancer research, maintenance of nuclear weapons, climate research, and astrophysics). The development of CSE software has some unique constraints that make it a good example of a non-traditional software development environment. To address the growing interest and need for high-quality computational science software, the goal of this proposal is to develop and deliver curriculum modules focused specifically on software for large scale computational science. In this work, the PI will develop a curriculum consisting of software engineering modules tailored for use by computational scientist developers. These modules will be designed to educate non-traditional developers about relevant software engineering tools and techniques. They will be packaged for use either in the university environment, embedded in existing courses or as part of a new course, or in the field, as training courses offered at various national labs and appropriate conference venues like the international meeting, Super Computing. The work uses a case study approach coupled with an interdisciplinary workshop. The plan is to cover the topics of validation and verification and agile methods by focusing on test-driven development and software inspection as well as the general problem of requirements gathering coupled with discovery.
This work will result in packaged, reusable SE modules created specifically for nontraditional developers. These modules will address the most pressing needs of CSE developers. The modules will be delivered in a variety of venues and the material evaluated during use, as well as measures of participant learning. Providing support for the development of high-quality CSE software will have broad impacts in advancing scientific and engineering knowledge in many domains.
