Article Preview
TopIntroduction And Background
The development of software applications for use in Computational Science and Engineering (CSE) is an important and growing segment of the software development landscape. The development of CSE software differs from the development of the typical types of software that software engineering researchers study (Carver, Kendall, Squires, & Post, 2007; Carver, 2009; Kelly, 2007; Segal, 2008; Wilson, 2006). These differences mean that many existing software engineering practices are not directly applicable within the CSE domain. Some of these characteristics are:
- •
The requirements discovery and gathering process is often difficult due to the large number of unknowns at the beginning of a project;
- •
Achieving the correct scientific or engineering output is more important to the developers than using correct or appropriate software engineering techniques; and
- •
Even when they are aware of traditional software engineering processes, which is not common, developers tend to be averse to using those processes.
Most software engineering research has been conducted in a business-oriented environment rather than a research-oriented environment. It is not clear whether the results of this research and the lessons learned about the usefulness of various techniques will translate to the CSE environment. To better understand the impact of the CSE environment and to document successful CSE software development practices, the author, along with other colleagues, as part of the DARPA High Productivity Computing Systems (HPCS) project (http://www.highproductivity.org), have conducted a series of case studies of successful CSE projects. To protect the anonymity of the projects, we assign each one a bird pseudonym. This paper describes the seventh case study in the series after Falcon (Post, Kendall, & Whitney, 2005), Hawk (Kendall et al., 2005), Condor (Kendall, Mark, Post, Squires, & Halverson, 2005), Eagle (Kendall, Post, Squires, & Carver, 2006), Nene (Kendall, Post, & Mark, January, 2007), and Osprey (Kendall et al., 2008) which will be referred to as the HARRIER project.
The common objectives for all of the case studies in the series are to (Carver et al., 2007):
- •
provide feedback to software vendors about issues that should be addressed to improve productivity;
- •
develop a body of case studies of CSE projects that can serve as a guide to the community; and
- •
document lessons learned about CSE software development to provide a resource for other similar teams.
The case study approach, as defined by Yin, is a well-understood method for gathering this type of information (Yin, 2002). The case studies, including the one reported in this paper, all followed a similar format that has proved successful. To provide context for this paper, a brief summary of the methodology is provided. The complete methodology is described in a previous publication (Carver et al., 2007). After identifying a suitable project, members of the code team complete a survey to provide background information about the project. Using this information, the researcher develops a set of questions that will be used to explore interesting topics further during an on-site interview. The researcher conducts the on-site interview with the team to gather more detailed information about the software development processes. Finally, using information from the survey and interview, the researcher writes a report that is circulated with the development team to ensure its accuracy.