Towards strongDynamicstrong Adaptation of Probabilistic strongSystemsstrong.pdfVIP

Towards Dynamic Adaptation of Probabilistic Systems 1 2 1 S. Andova , L.P.J. Groenewegen and E.P. de Vink 1 Formal Methods, TU Eindhoven, The Netherlands 2 FaST Group, LIACS, Leiden University, The Netherlands Abstract. Dynamic system adaptation is modeled in the coordination language Paradigm as coordination of collaborating components. A spe- cial component McPal allows for addition of new behavior, of new con- straints and of new control in view of a new collaboration. McPal grad- ually adapts the system dynamics. It is shown that the approach also applies to the probabilistic setting. For a client-server example, where McPal adds, step-by-step, probabilistic behavior to deterministic com- ponents, precise modeling of changing system dynamics is achieved. This modeling of the transient behavior, spanning the complete migration range from as-is collaboration to to-be collaboration, serves as a step- ping stone to quantitative analysis of the system during adaptation. 1 Introduction Many systems today are affected while running by changes in their operational environment, while they cannot be shutdown to be updated and restarted again. Instead, dynamic adaptive systems must be able to manage adaptation steps on- the-fly to accommodate a new plan. Dynamic adaptive systems usually consist of interactive components, architecturally organized. However, system adaptation requires proper coordination. A carefully chosen coordination should guarantee that, during the adaptation, the system functionality is neither interrupted nor disturbed, and non-functional quantities, though possibly changing, should not exceed allowed boun