Autonomic systems encompass elements, components, and sometimes smaller autonomic sub-systems. Composition, integration and orchestration at different levels of such autonomic systems are significant issues in the interoperability of their constituent elements. This research focuses on orchestration as an important process of interoperability in an abstract autonomic system model. Policybased management has potential to play an important role in orchestrating constituent components of autonomic systems/ elements, and also to fill the gap between business and IT objectives. This research applies policy-based orchestration using crisp and fuzzy policies to reach the business goals of the system, and to obey the constraints defined by the Service Level Agreement (SLA). First, a simulated model has been used in MATLAB/Simulink environment, and then the system is implemented by Agent Building and Learning Environment (ABLE) in Java.