Victor Basili

Software engineering is not only an increasingly challenging endeavor that goes beyond the intellectual capabilities of any single individual engineer, but is also an intensely human one. Tools and methods to develop software are employed by engineers of varied backgrounds within a large variety of organizations and application domains. As a result, the variation in challenges and practices in system requirements, architecture, and quality assurance is staggering. Human, domain and organizational factors define the context within which software engineering methodologies and technologies are to be applied and therefore the context that research needs to account for, if it is to be impactful. This paper provides an assessment of the current challenges faced by software engineering research in achieving its potential, a description of the root causes of such challenges, and a proposal for the field to move forward and become more impactful through collaborative research and innovation between public research and industry.

Enterprise Resource Planning ERP systems integrate information across an entire organization that automate core activities such as finance accounting, human resources, manufacturing, production and supply chain management etc. to facilitate an integrated centralized system and rapid decision making resulting in cost reduction, greater planning, and increased control. Many organizations are updating their current management information systems with ERP systems. This is not a trivial task. They have to identify the organizations objectives and satisfy a myriad of stakeholders. They have to understand what business processes they have, how they can be improved, and what particular systems would best suit their needs. They have to understand how an ERP system is built, it involves the modification of an existing system with its own set of business rules. Deciding what to ask for and how to select the best option is a very complex operation and there is limited experience with this type of contracting in organizations. In this paper we discuss a particular experience with contracting out an ERP system, provide some lessons learned, and offer suggestions in how the RFP and bid selection processes could have been improved.

Cost estimation is a critical issue for software organizations. Good estimates can help us make more informed decisions (controlling and planning software risks), if they are reliable (correct) and valid (stable). In this study, we apply a variable reduction technique (based on auto-associative feed-forward neural networks-called Curvilinear component analysis) to log-linear regression functions calibrated with ordinary least squares. Based on a COCOMO 81 data set, we show that Curvilinear component analysis can improve the estimation model accuracy by turning the initial input variables into an equivalent and more compact representation. We show that, the models obtained by applying Curvilinear component analysis are more parsimonious, correct, and reliable.

In order to understand how high performance computing (HPC) programs are developed, a series of experiments, using students in graduate level HPC classes and various research centers, were conducted at various locations in the US. In this report, we discuss this research, give some of the early results of those experiments, and describe a web-based Experiment Manager we are developing that allows us to run studies more easily and consistently at universities and laboratories, allowing us to generate results that more accurately reflect the process of building HPC programs.