Control of research oriented software development

The Nuclear Waste Policy Act of 1982 directs the Department of Energy (DOE) to dispose permanently high level radioactive waste and civilian spent nuclear fuel by January 31, 1998. DOE has responded by creating an organizational structure that directs all the activities necessary to carry out the legislative demands. LLNL is conducting research in the earth sciences and is developing some unique computer codes to help establish the feasibility of geologic repositories for nuclear waste. LLNL has several codes under development. This paper examines the administrative and organizational measures that were and still are being undertaken in order to control the development of the two major codes. In the case of one code, the software quality assurance requirements were imposed five years after the code began its development. This required a retroactive application of requirements. The other code is still in the conceptual stages of development and here requirements can be applied as soon as the initial code design begins. Both codes are being developed by scientists, not computer programmers, and both are modeling codes, not data acquisition and reduction codes. Also the projects for which these codes are being developed have slightly different software quality assurance requirements. All these factors contribute unique difficulties in attempts to assure that the development not only results in a reliable prediction, but that whatever the reliability, it can be objectively shown to exist. The paper will examine a software management model. It will also discuss the reasons why it is felt that this particular model would stand a reasonable chance for success. The paper will then describe the way in which the model should be integrated into the existing management configuration and tradition