NASA Application of SEER-SEM
Those that know the full history of Galorath know that the Jet Propulsion Laboratory was the sponsor of Dan Galorath’s first work in software estimating in the early ‘80s.
Application of the primitive model resulting from the JPL study found that such modeling can help make software projects successful.
When Mr. Galorath was brought in to rescue the project (a commercial embedded system), the hardware people initially declared that the software team was incompetent and would need to be replaced. Mr. Galorath ran that model and found the issue was the memory sizing in the hardware itself. He went to engineering and asked for more memory. He was told to quit making excuses. Taking that model to management got the memory increase approved… Management by quantification rather than gut feel. The project was then completed on a timely basis with substantial reduction of software development cost. That showed him that such models could be viable in government and industry. SEER is many generations better than that early model. And it continues to make projects successful even today. Of course today there is SEER for Hardware, Electronics & Systems, and more that could estimate the entire program.
These many years later Galorath is proud to be a part of NASA and NASA’s success with the SEER Models.
Several years ago the NASA cots symposium included a paper by Sherry Stukes and John Spagnuolo, Jr on the benefit of using SEER for Software (SEER-SEM) for decisions and knowledge engineering. The paper abstract includes:
“At The California Institute of Technology/Jet Propulsion Laboratory, methodologies for Flight Software (FSW) cost estimation and documentation are determined that allow for efficient concurrent and consistent analysis within a tight schedule constraint. This knowledge is structured or “engineered” to facilitate the implementation of FSW cost estimation by others who wish to serve as practitioners in the field. Knowledge Engineering (as defined by Edward Feigenbaum and Pamela McCorduck in 19831 ) “… is that discipline that involves integrating knowledge into computer systems in order to solve complex problems normally requiring a high level of human expertise”. Embedded in this definition is the acquisition and structuring of the related information characterizing the knowledge domain of interest.
The effort described in this presentation relates to these ideas in 2 specific ways: (1) It gives a decision graph relating to the acquisition, structuring and representation of the knowledge used for the computation of FSW estimates for space missions at JPL and (2) Although we do not fully automate the processes described here, various aspects of the work are embedded in and related to computer activity. Further, the work is done in such a way as to facilitate further automation of its procedures. We present an overview of FSW cost estimation techniques used for Independent Cost Estimates (ICEs), proposals, and for the validation of Cost Analysis Data Requirements (CADRe) submissions. The aforementioned decision graph illustrates the steps taken in the production of these estimated costs and serves as the basis of discussion for this paper and the corresponding presentation. General principles for the estimation of FSW are presented using the SEER-SEM computer program as an illustration of these principles when appropriate. A discussion of various Source Lines of Code (SLOC) data sources and their uses for the preparation of the estimates is given as code size is a major driver for software costs. A computerized methodology used to map the SEER-SEM output into the JPL Work Breakdown Structure (WBS) is illustrated. Finally, an “Across the Board” tally of the SEER-SEM runs and their corresponding input is given in a single sheet for a set of several proposals at JPL. This paper is not a description per se of the efforts by two software cost analysts.Go Back