Model-driven Architecture for Inter-component Dependency, Fault-tolerance, and Execution Environment Reconfigurability

Large-scale engineering systems require very large real-time computation. The computers that effect on-line processing must be extremely reliable, because of the large cost of operating the facilities and because of the potential loss of data or inability to deliver critical functionality in response to real-time threats. Standard, redundant fault tolerant strategies and mechanisms are inappropriate for such systems, because of the associated overheads implied (10% is considered a critical maximum overhead). Adaptive, fault-mitigation approaches that achieve the 10% overhead limit must be used instead. In this proposal, a set of tools for specifying and implementing user-defined adaptive algorithms based on model-driven architecture of both software and hardware subsystem specification, is proposed. We describe a hierarchical fault-tolerance strategy, introduce model-integrated computing tool for hierarchical inter-component system design and implementation, and propose design space exploration for dynamically reconfigurable execution environment. In Phase I, this concept will be refined and its validity for full prototyping in Phase II will be demonstrated.