Metric-driven analysis and feedback systems for enabling empirically guided software development
Richard W. Selby, Adam Porter, Doug Schmidt, Jim Berney · 1991
Metric-driven analysis and feedback systems enable developers to define empirically guided processes for software development and maintenance. These sys-tems enable developers to achieve S. E.I. process matu-rit y levels 4 and 5, “managed ” and “optimizing ” pro-cesses. We define a set of architectural principles and abstract interfaces for designing metric-driven analysis and feedback systems. A fundamental principle is to make measurement active by integrating measurement and process, which contrasts with the primarily pas-sive use of measurement in the past. A second princi-ple is that these systems should provide capabilities for user-specifiable interpretation of process events, object state changes, and calendar time abstractions in order to allow a representative basis for project analysis. An-other principle is that these systems should provide an extensible integration framework that supports the ad-dition of new techniques and integrates approaches for synergistic application. These and other architectural principles are em-bodied in a prototype analysis and feedback system, called Amadeus. This prototype system is intended to demonstrate the feasibility and merit of these systems and the principles for designing them. An initial ver-sion of the system is running, and it has been used in several evaluative studies. We define the abstract in-terfaces in Amadeus and describe its operation. Metric-driven analysis and feedback systems influence the de-sign of process modeling and programming languages and software environment architect ures. According] y,