Improving the Component Discovery Process by Leveraging Automatic Sensitive Analysis

Component-based approaches have acquired a prominent role in development of complex software systems. Successful reuse of existing components requires being able to first identify, and then distinguish among, functionally (near-) equivalent elements of large component collections. Similar components can be ranked using quality criteria, thus, some goal-oriented techniques attempt to quantify components quality by indicating valid ranges for their properties and behavior, like stability, latency and so on. Unfortunately, most current techniques yield non-robust ranges, and most tools do not allow architects to observe the range selection during the process. This paper presents a technique for sensitivity analysis of components discovery, built over fuzzy sets. A prototypical tool has been built, and use of the technique and tool are illustrated with an example. This iterative approach allows evaluators to compare "what if" scenarios for alternative component quality criteria, supporting requirements evolution without continuous expert support to recalibrate valid property ranges.