Caracterização de software científico: um estudo de caso em modelagem computacional

Abstract

In the Internet era, scientific software are not only an instrument to generate results but also crucial to most researches. A lot of scientists, however, learn what they know about software development informally, and this informality typically generates a random character to the final software products, making it difficult to reuse and interoperate. Scientific software specifically developed to meet individual scientists needs in their respective areas of work have become very important both in its development process as in its use and, depending on the nature of scientific activity, it is essential. Another feature of scientific computing is that the scientific software are also often exploited, developed with the focus of helping to understand a problem. Many software engineering disciplines can contribute to the generation of high-quality scientific products. The application of models, techniques and tools contribute to the construction of scientific software more reliable, secure and efficient and with fewer failures during its execution. Recent studies point to the application of agile methods, because they are more receptive to change and deal better with emergency requirements. This research aims to characterize the development of scientific software, through an exploratory study with a group of researchers from LNCC, experts in computational modeling area. For the characterization of scientific software we built an instrument with the following research questions: identify scientific software quality criteria, identification of development process models for scientific software, adoption possibilities for Software Engineering practices, possibility of scientific software development by specialized team instead of scientists themselves, re-use of scientific artifacts, sharing of scientific software and identification of scientific workflows characteristics. After the application of the instrument, the results were quantified and analyzed, identifying characteristics of scientific software and scientific workflow. At the end we define some proposals for software engineering practices adoption by scientists involved in this research. The results provide evidences on how to improve scientific software, its development process, sharing and reuse.