Author

Rongshu Yi

Date of Award

4-2007

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Computing and Software

Supervisor

Tom Maibaum

Language

English

Abstract

As one of the main objectives of Enterprise Business Applications, high availability is becoming more and more important and is a widespread concern of customers, architects and developers. But, until now there are still no models for computing and evaluating accurately the value of a system's availability, and no any feasible and systematic methods to improve the availability 'of a software system, which may be made up of several patterns, and may be very complex. PBSA(Part Based Software Architecture), as a new analytical model for predicting a complex software system's availability, is based on the availability of each component, the properties of some main architecture patterns, and the architectural structure of the software system itself. It uses a Markov model to translate the evaluation process of a system's availability into a Markov Transfer process, and uses the properties of Markov Chains to calculate and predict the probability of the whole system's availability. Then based on the model, a sensitivity rank for architecture components and factors can be given, and systematic strategies for improving the availability property of a complex application system can be given. As one of the strategies for improving system availability, a new component, named the HA(high availability) manager component, is defined and introduced, together with some idea of an HA Hierarchy.

McMaster University Library

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS