Dependability Aspects in Configurable Embedded Operating Systems (DanceOS)

SPP 1500: Entwurf und Architekturen verlässlicher Eingebetteter Systeme

Future hardware designs for embedded systems will exhibit more parallelism at the price of being less reliable. This bears new challenges for system software, especially the operating system, which has to use and provide software measures to compensate for unreliable hardware. However, dependability in this respect is a nonfunctional concern that affects and depends on all parts of the system. Tackling it in a problem-oriented way by the operating system is an open challenge: (1) It is still unclear, which combination of software measures is most beneficial to compensate certain hardware failures – ideally these measures should be understood as a matter of configuration and adaptation. (2) To achieve overall dependability, the implementation of these measures, even though provided by the operating system, cannot be scoped just to the operating-system layer – it inherently crosscuts the whole software stack. (3) To achieve cost-efficiency with respect to hardware and energy, the measures have, furthermore, to be tailored with respect to the actual hardware properties and reliability requirements of the application. We address these challenges for operating-system design by a novel combination of (1) speculative and resource-efficient fault-tolerance techniques, which can (2) flexibly be applied to the operating system and the application by means of aspect-oriented programming, driven by (3) a tool-based (semi-)automatic analysis of the application and operating-system code, resulting in a strictly problem-oriented tailoring of the latter with respect to hardware-fault tolerance.