Testability is an important quality attribute of software, especially for critical systems such as avionics, medical, and automotive. Improvement in the early testability of software architecture, the first artifact of the software system, will help reduce issues and costs later in the development process.
AADL, an architecture analysis description language suitable for critical embedded, real-time systems, can be used for design documentation, analysis and code generation. Because the capability of AADL can be extended, it is possible to add new analyses to its core language. Tools such as the Open Source AADL Tool Environment (OSATE) provide plugins for processing AADL models. Although adding new plugins in OSATE extends AADL, there currently exists no AADL extension for testability measurement. The purpose of this thesis is to propose such a method to measure the testability of AADL models as well as to develop a testability plugin in OSATE.
Much research has been conducted on testability of hardware, software and embedded systems, resulting in several approaches for measuring this quality attribute. Among them, the approach measuring testability as a product of controllability and observability using information transfer graph (ITG) is the most applicable for measuring the testability of AADL models. This thesis proposes a method applying this approach to AADL models. A complete testability measure plugin for OSATE was developed based on this approach and detailed examples are given in this thesis to demonstrate its applicability.