Realising Adaptive Web Services through Automated Policy Refinement

Abstract

Traditionally policy based management systems have relied on 'engineered' policy refinement to decompose high level policies down to low level policies for implementation 111121. Such management systems typically rely on a policy designer ensuring that a high level policy creates the necessary invocations or events to trigger the correct low level policies (and their implementation). The difficulty with this approach is that human error in the policy design can lead to feature interaction problems, conflicts and errors 131, 141. This refinement problem is a key challenge for policy based management to be adopted widely within service management (voted #1 key challenge at Policy 2005 panel session). In this paper we concentrate on policy based management of web services, where the web services have adaptive behaviors. The paper proposes the design, implementation and evaluation of a suite of policy development tools which (i) facilitate the specification of adaptive web services (which are to be managed by a policy based system) using finite state machines to describe the permitted adaptive behaviors, (ii) supports the specification of supportable (functionally correct) high level management policies, the automated refinement of these high level policies and the auto-generation of enforceable (low level) policies. The paper also presents an evaluation of the systems with particular reference to the usability of the proposed solution. The paper also provides a brief presentation of related research in policy refinement and distinguishes the proposed approach from published work. Finally the paper concludes by drawing initial findings and identifies future work.