Return to search

DRACA: Decision-support for Root Cause Analysis and Change Impact Analysis

Most companies relying on an Information Technology (IT) system for their
daily operations heavily invest in its maintenance. Tools that monitor network
traffic, record anomalies and keep track of the changes that occur in the system
are usually used. Root cause analysis and change impact analysis are two main
activities involved in the management of IT systems. Currently, there exists no
universal model to guide analysts while performing these activities. Although the
Information Technology Infrastructure Library (ITIL) provides a guide to the or-
ganization and structure of the tools and processes used to manage IT systems, it
does not provide any models that can be used to implement the required features.
This thesis focuses on providing simple and effective models and processes for
root cause analysis and change impact analysis through mining useful artifacts
stored in a Confguration Management Database (CMDB). The CMDB contains
information about the different components in a system, called Confguration Items
(CIs), as well as the relationships between them. Change reports and incident
reports are also stored in a CMDB. The result of our work is the Decision support
for Root cause Analysis and Change impact Analysis (DRACA) framework which
suggests possible root cause(s) of a problem, as well as possible CIs involved in a change set based on di erent proposed models. The contributions of this thesis are
as follows:

- An exploration of data repositories (CMDBs) that have not been previously
attempted in the mining software repositories research community.

- A causality model providing decision support for root cause analysis based
on this mined data.

- A process for mining historical change information to suggest CIs for future
change sets based on a ranking model. Support and con dence measures are
used to make the suggestions.

- Empirical results from applying the proposed change impact analysis process
to industrial data. Our results show that the change sets in the CMDB were
highly predictive, and that with a confidence threshold of 80% and a half
life of 12 months, an overall recall of 69.8% and a precision of 88.5% were
achieved.

- An overview of lessons learned from using a CMDB, and the observations we
made while working with the CMDB.

Identiferoai:union.ndltd.org:WATERLOO/oai:uwspace.uwaterloo.ca:10012/4889
Date12 1900
CreatorsNadi, Sarah
Source SetsUniversity of Waterloo Electronic Theses Repository
LanguageEnglish
Detected LanguageEnglish
TypeThesis or Dissertation

Page generated in 0.0016 seconds