A Policy-based Management Framework for Pervasive Systems using Axiomatized Rule-Actions

Abstract

Pervasive systems comprise large collections of heterogeneous and mobile devices, services and applications. A management infrastructure is required to govern the system behavior according to policies specified by the system administrator. Policy-based management is a well-established approach where policies are specified as Event-Condition-Action (ECA) rules that determine the management actions to be performed when certain situations occur. The problem with ECA policies is that conflicting actions may get triggered on the same event resulting in policy conflicts. Cycles may result when a set of policy rules trigger each other continuously. Existing approaches to conflict detection are limited in scope and can only detect conflicting actions if they are explicitly stated. In addition, current techniques do not detect cycles in management policies. We propose an extension to the ECA rule framework, called Event-Condition-PreCondition-Action-PostCondition (ECPAP) as a rule framework for management policies. In this framework, actions are annotated with axiomatic specifications that enable powerful reasoning to detect conflicts and cycles in policies. We present the details of this framework