Dependability of Component Based Systems

Summary

It is now commonplace to develop software based systems from components (e.g. these may be so called commercial off the shelf components, the results of an object oriented development, the evolution of existing product lines). The goal is to describe, design or select components and then assemble large systems according to architectural principles. Approaches are often sought that minimise the need to know implementation details of the components and to rely on specification of the interface behaviour.

There is usually uncertainty in the evidence that would support claims of dependability of the components. But such evidence is indispensable for critical applications such as medical, aerospace, automobile, financial applications in national infrastructure and embedded systems in the home. Another trend is the proliferation of applications where dependability of software is critical. For such applications

• Dependability-related attributes of components whose implementation details are not known or are uncertain must be assessed,
• The overall system attributes (functionality, reliability, robustness etc.) must be translated into requirements for components or synthesised from the component attributes,
• Techniques are needed that can guarantee or at least assure certain dependability-related properties of a system even it is assembled of components for which no guarantee is given.

To tackle these problems an interdisciplinary approach is needed that combines safety and requirements analysis techniques, specification techniques, design adaptation techniques such as wrappers and adapters and probabilistic modelling of decision making under uncertainty.

The integration of disparate sources of evidence is another challenge of component-based dependable systems.

The 20 talks of the seminar covered the following topics (among others):

• system specification and generating specifications from requirements,
• modular certification,
• justification of safety,
• fault tolerance,
• evolution of systems,
• modelling of systems by Abstract State Machines and Petri Nets,
• synthesizing refinements,
• test automation.

The seminar brought together researchers and practitioners in order to achieve a common understanding of the problems and collect possible solutions. We experienced synergetic effects by inter-disciplinary working.

Besides the technical aspects of safety and component-orientation, questions of certification and standardisation were discussed. The week was structured to facilitate industrial involvement.