Environmental Engineering Reference
In-Depth Information
Engine essentially provides server-level correlation of multiple IT systems to
assist with root-cause analysis and automated corrective action [
41
]. The
ABLE rules engine can be used for more complex analysis. In effect, it is
an agent-building learning environment that includes time series analysis
and Bayes classification among others. It correlates events and invokes the
necessary action policy [
41
].
It has been noted that correlation, rule discovery, and root-cause analysis
activities can benefit from incorporating Bayesian Networks [
153
], either in
the rule discovery process or in the actual model learning to assist with self-
healing [
150
]. Large-scale server management and control has also received
similar treatment. Event logs from a 250-node large-scale server were analyzed
by applying a number of machine-learning algorithms and AI techniques to
establish time-series methods, rule-based classification, and BN algorithms for
a self-management and control system [
129
].
Another aspect of monitoring and root-cause analysis is the calculation
of costs, in conjunction with the self-healing equation in an autonomic sys-
tem. One approach utilizes naive Bayes for cost-sensitive classification and a
feedback approach based on a Markov decision process for failure remedia-
tion [
89
]. The argument is easily made that the autonomic system involves
decisions and decisions involve costs [
25
]. This naturally leads to work with
agents, incentives, costs, and competition for resource allocation and exten-
sions thereof [
25
,
105
].
8.2.5 Legacy Systems and Autonomic Environments
Autonomic Systems is arguably widely believed to be a promising approach
to develop new systems. Yet organizations continue to have to deal with the
reality of either legacy systems or building “systems of systems” composed of
new and legacy components that involve disparate technologies from numer-
ous vendors [
76
]. Work is currently underway to add autonomic capabilities
to legacy systems in areas such as instant messaging, spam detection, load
balancing, and middleware software [
76
].
Generally, the engineering of autonomic capabilities into legacy systems
involves providing an environment for monitoring the system's sensors and
providing adjustments through effectors to create a control loop. One such
infrastructure is Kinesthetics eXtreme (KX). It runs a lightweight, decen-
tralized, easily integratable collection of active middleware components tied
together via a publish-subscribe (content-based messaging) event system [
76
].
Another tool, called Astrolabe, may be used to automate self-configuration
and monitoring and control adaptation [
14
]. The AutoMate project, in-
corporating ACCORD (an autonomic component framework), utilizes the
distributed interactive object substrate (DIOS) environment to provide mech-
anisms to directly enhance traditional computational objects/components
