Home // International Journal On Advances in Intelligent Systems, volume 2, number 4, 2009 // View article

Model Transformations Given Policy Modifications in Autonomic Management

Authors:
Raphael M. Bahati
Michael A. Bauer

Keywords: Model Transformation, Model Adaptation, Reinforcement Learning, Autonomic Management, Policy-based Management.

Abstract:
This paper presents an approach for adapting a model learned from the use of an active set of policies to run-time policy modifications. The work leverages our most resent efforts utilizing Reinforcement Learning methodologies to facilitate dynamic use of policies within autonomic computing. The use of policies in this context offers significant benefits to autonomic systems in that it allows systems to determine how to effectively meet the desired, and at times seemingly conflicting, objectives under dynamically changing conditions. Contrary to other approaches that make use of some form of learning to model performance management, our approach enables the model learnt from the use of an active set of policies to be reused once those policies change. Since the learning mechanisms are modelled from the structure of the policies, policy modifications can be mapped onto the learnt model. Our analysis of the policy modifications suggest that most of the learned model could be reused, potentially accelerating the learning process. In this paper, we provide formal definitions on the different kinds of policy modifications that might occur as well as elaborate, with detailed examples, on how such modifications could impact the currently learned model.

Pages: 477 to 497

Copyright: Copyright (c) to authors, 2009. Used with permission.

Publication date: March 17, 2010

Published in: journal

ISSN: 1942-2679