Toward self-reconfiguring, fault-adaptive, high-performance distributed real-time systems

Designing embedded systems that guarantee some form of reasonable behavior in the presence of failure is a difficult task, and the addition of scalability and real-time constraints only serve to exacerbate the problem. In large-scale systems such as those used in high-energy physics experiments, the necessary conditions to ensure complete and reliable operation of the system can not be guaranteed. Moreover, typical methods of fault tolerance can not be applied to these systems given their size and budgetary constraints. As the performance, size, and inherent complexity of these systems increase, it becomes necessary to develop increasingly advanced tools and techniques to help manage their design, implementation, and configuration. This paper outlines an approach toward designing large-scale embedded systems by leveraging concepts of Model Integrated Computing to configure a scalable Reflex and Healing architecture to implement model-based user-defined fault-mitigation strategies within the system.