Model based reactive planning and prediction for autonomic systems
Peter H. Deussen · Autonomic Computing and Communication Systems · 2007
Dynamic adaptation according to situational variety requires the employment of planning algorithms for reactive and proactive dealing with detected or predicted conditions. The complexity of large scaled, distributed systems however prohibits planning strategies acting on the granularity of single components because (a) a local system view from the perspective of singular components is usually not sufficient, and (b) the synthesis (or a-priory definition) of a global perspective is far beyond feasibility. We therefore propose the use of dynamic abstraction mechanisms to generate planning models of a suitable degree of granularity. The framework of Abstract Interpretation is used to define both a reactive planning algorithm and a pro-active prediction.