National Institute of Statistical Sciences
19 T. W. Alexander Drive
P.O. Box 14006
Research Triangle Park, NC 27709-4006
Tel: 919.685.9300
FAX: 919.685.9310
admin@niss.org

Code decay in Legacy Software Systems:
Measurement, Models and Statistical Strategies

 

Project Information


PIs Alan F. Karr, NISS; Stephen G. Eick, Bell Labs, Adam A. Porter, University of Maryland (College Park)
Source of Funds
NSF Program on Transformation to Quality Organizations
Matching funds from Lucent Technologies and NISS
Funding Level $81,000 from NSF in FY 1998
Dates April 15, 1996 - March 31, 1999

Project Description

This project is a partnership with Bell Laboratories (now part of Lucent Technologies). It is directed at improving the development and maintenance processes for large legacy software systems by application of statistical strategies to model and control the effects of design decisions, software architecture, and organizational factors. Improvements will be reductions in cost and development time, increases in software quality and improved capability to predict cost, schedule and quality.

Decay of computer software is a problem of massive but largely unappreciated proportions. A timely illustration is recent (Summer, 1995) problems with the Federal Aviation Administration's (FAA) air traffic control system. Failing 1960s vintage hardware must be replaced, but the software will not run on more modern (even 1970s) hardware, and it is seen as so old, inaccessible and frail - so decayed - that no one dares modify it. Instead, unnecessarily expensive and grossly inefficient short-term alternatives must be implemented.

The research program is directed toward three key problems regarding code decay:

While code decay can be described conceptually, precise definitions admitting quantification are less clear. We developing indices (measures) of code decay that reflect key behavioral and managerial characteristics in the software development process, and are inked closely to the history and structure of the code. These code decay indices serve as scientific "pivots" in that they can be changed by changing controllable (e.g., organizational) factors, and that they can be used to change key response variables: cost, interval and quality. The indices constitute a key intermediary on the path from factors to responses.

Research in the first six months of the project has focused on extraction from a much larger data base of the change history of the code for one subsystem of the 5ESS® telephone switch; on development of World Wide Web-based tools for extraction and presentation of descriptive statistics (such as sizes and temporal patterns of changes); on preliminary models for changes to software to induce faults and initial experiments targeted at inducing accelerated code decay.

Additional Information

 

NISS Home Page
Help