Browse Prior Art Database

A Autonomic Policy Mechanism

IP.com Disclosure Number: IPCOM000033519D
Original Publication Date: 2004-Dec-13
Included in the Prior Art Database: 2004-Dec-13
Document File: 2 page(s) / 32K

Publishing Venue

IBM

Abstract

By definition, autonomic systems require no ongoing human interaction. That does not imply that humans are uninvolved. Autonomic systems are governed by policies that are defined by humans. Such policies can be defined by system developers setting defaults policies or by system owners. Once policies are defined, autonomic systems must operate according to those policies. We disclose an autonomic policy mechanism.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

A Autonomic Policy Mechanism

By definition, autonomic systems require no ongoing human interaction. That does not imply that humans are uninvolved. Autonomic systems are governed by policies that are defined by humans. Such policies can be defined by system developers setting defaults policies or by system owners. Once policies are defined, autonomic systems must operate according to those policies. We disclose an autonomic policy mechanism.

We assume that policy specification contain 4 component: scope, condition, business value and measurable intent.

We disclose a 5-phase policy mechanism that use these components. The 5 phases are: definition, scoping, condition evaluation, value optimization and execution. In the definition phase, policy definers input a set of policies. Scope is the assignment of policies to the appropriate entities. ("Entity" is an inclusive term covering resources, services, etc.) Entities perform conditional evaluation to determine whether a policy should be active. Using value optimization, entities determine which of the active, and perhaps conflicting, policies should be enforced. Execution of the policy follows. Policies can be chained so the outcome of one policy determines which policies are subsequently active.

Policies are adaptable. Autonomic systems can adjust policy charactistics, typically conditions and values, in response to observed outcomes of previous applications of the policy.

Policies are layered to ensure that, under all conditions, the system knows how to operate. The root policy, or union of root policies, must cover all situations, not just all anticipated situations. For example, a root policy might state that if the system does not know how to respond to its state for some interval of time, then the system should be restarted.

The five phases are relatively straightforward in isolation, yet in combination, meet the broad requirements of autonomic systems. We describe in outline the five phases below.

Definition: an adminstrator sets the system policies. This can be done either using a policy definition tool or by writing the policies with a text editor. In either case, policies are converted i...