A MIC based image processing system has been
described which supports design and implementation of autonomic/self-correcting or adapting
behaviors. These systems are guided through their adaptation process by user specifications placed
into the model. The specifications permit a hybrid implementation of classical controls-based
approaches for parametric adaptation, along with structural adaptation for managing gross changes
in system performance. A domain-specific language is described based on classical control
mechanisms such as probes/monitors, evaluators, and control adaptation structures. These are in the
early stages of a formal language specification.
Early versions of the language have been implemented in the MIC/GME infrastructure from Vanderbilt
University. Representative portions of an example of a system model were presented, showing the
specification of computations and adaptations. Performance is yet to be evaluated, however we are
encouraged by prior work using automatic behavior for fault tolerance in large-scale parallel
systems
for high-energy physics processing.