Autonomic Systems - Autonomous and Autonomic Systems - page 175

element of autonomic computing [ 128 ]. This work includes probabilistic rea-
soning, and prospectively, should be able to benefit from invoking genetic
algorithms for model selection.
Probabilistic techniques such as Bayesian networks (BNs) discussed in [ 50 ]
are also central in research into autonomic algorithm selection, along with
self-training and self-optimizing [ 50 ]. Re-optimization of enterprise business
objectives [ 4 ] can be encompassed by the breadth and scope of the autonomic
vision through such far-reaching work combined with AI techniques (machine
learning, Tabu search, statistical reasoning, and clustering analysis).
As an example, the application “Smart Doorplates” assists visitors to a
building by locating individuals who are not in their oces. A module in the
architecture utilizes probabilistic reasoning to predict the next location of an
individual, which is reported along with his/her current location [ 173 , 174 ].
8.2.3 Knowledge Capture and Representation
Vital to the success of Autonomic Systems is the ability to transfer expert hu-
man knowledge about system management and configuration to the software
managing the system. Fundamentally, this is a knowledge-acquisition prob-
lem [ 85 ]. One current research approach is to capture the expert's actions
automatically (keystrokes and mouse movements, etc.) when performing on
a live system, and dynamically build a procedure model that can execute
on a new system and repeat the same task [ 85 ]. Establishing a collection of
traces over time should allow the approach to develop a generic and adaptive
model.
The Tivoli management environment approaches this problem by captur-
ing in its resource model the key characteristics of a managed resource [ 77 ].
This approach is being extended to capture the best practices information into
the common information model (CIM), through descriptive logics at both the
design phase and the deployment phase of the development lifecycle [ 83 ]. In
effect, the approach captures system knowledge from the creators, ultimately
to perform automated reasoning when managing the system.
8.2.4 Monitoring and Root-Cause Analysis
Event correlation, rule development, and root-cause analysis are important
functions for an autonomic system [ 155 ]. Early versions of tools and autonomic
functionality updates to existing tools and software suites in this area have
recently been released by IBM [ 41 ] through their AlphaWorks Autonomic
Zone website. Examples include the Log and Trace Tool, the Tivoli Autonomic
Monitoring Engine, and the ABLE rules engine.
The generic Log and Trace Tool correlates event logs from legacy systems
to identify patterns. These patterns can then be used to facilitate automa-
tion or support debugging efforts [ 41 ]. The Tivoli Autonomic Monitoring