Qualification and certification of software-reliant safety-critical cyber-physical systems has increasingly become a challenge: 70% of embedded software system errors are introduced during requirements and architecture design, while 80% are found during system integration or later, resulting in exponentially growing rework and qualification costs. This is due to mismatched assumptions between the embedded system software components, the distributed hardware platform, and the physical system [read more]. Using traditional means, system integration becomes high risk, and system evolution (lifecycle support) becomes expensive and results in rapidly outdated components. System engineers do not have insight, until late in the process, into critical system characteristics such as
- performance (e.g., throughput or quality of service)
- time criticality
- fault tolerance
Advances in architecture research in the 1990s has led to an opportunity to address this industry problem through an architecture-centric model-based engineering approach. Aerospace industry sponsored the development of the Architecture Analysis & Design Language (AADL) under the auspices of SAE International. The standard was developed under technical leadership by the SEI and originally published in 2004, with revisions in 2009 and 2012, based on feedback from users.The AADL standard suite includes extensions to address specific needs for fault modeling, modeling of ARINC653 partitioned architectures, integration of data models, requirements, and code generation. The SEI also developed an open source tool set for AADL with analysis capabilities for multiple quality attribute dimensions.
To see more detailed information on AADL, click here.
System models that precisely capture this architecture provide the basis for predictable system engineering through repeated analysis early in and throughout the development life cycle with the following benefits:
- reducing risk through early and repeated analysis of the system architecture
- permitting the engineer to see system-wide impacts of architectural choices
- increasing confidence by validating model assumptions in the operational system (and permitting the system models to evolve in multiple fidelity)
- reducing cost through fewer system integration problems and simplified lifecycle support
Read more about the benefits of architecture-centric model-based engineering with AADL.
The value of SAE AADL as a modeling notation with well-defined semantics has been recognized by the research community, and AADL has become a technology platform of choice in integrating analytical research technologies (See the AADL Wiki and In the News).
Building on this concept of virtual integration, SEI research in this area is currently focusing on
- architecture fault modeling automating system safety and reliability analysis and validation to become a repeatable incremental process throughout the development lifecycle
- integration with formalized requirement tracking and evidence gathering through confidence maps
- understanding technical debt of architectural design decisions
- contract-based predictive validation of cyber-physical systems
To demonstrate its approach and technology, the SEI has developed a set of architecture-related tools for the design and analysis of software architectures. The SEI also publishes public models to demonstrate its approach for testing its approach and tools. Most of architecture models are published on our github examples area and described on our public wiki.
Also see Software Architecture Tools and Methods, Analyzing the Architecture.
Read about Virtual System Integration.