Context-aware autonomic systems in real-time applications: Performance evaluation

Nowadays one can observe rapid development of various technologies supporting autonomic features and context-awareness of pervasive systems. And this is perfectly understandable as these features effectively increase the systems autonomicity. However, in the case of safety-critical realtime systems it becomes crucial to determine how the entire system performance is affected by the growth in complexity of its autonomic / dynamic decision making engine which is processing the context information for control purposes. In this paper we consider policy-based computing as a candidate technology to support selected autonomic features and context-awareness in safety-critical systems. We conduct performance analysis in a generic policy-supervised system in order to determine the relation between policy complexity (reflecting decision making system complexity) and the decision evaluation time (affecting the entire system performance). We present a coherent characteristic polynomial method as a means of pre-estimation of policy evaluation time depending on its complexity. The results we present can be potentially used for pre-evaluation of whether a policy of a certain complexity level is of any applicability in a safety critical system that has to meet fixed real-time deadlines.