HELSINKI UNIVERSITY OF TECHNOLOGY

ABSTRACT OF MASTERíS THESIS

Author: Seppo Kauppinen

Title of thesis: Modeling Internet Service Production Ė Dependability and

Risk Assessment

Date: Aug 23rd 1999             Pages: 69+7

Department: Department of Engineering Physics and Mathematics

Professorship: Mat-2 Applied Mathematics

Supervisor: Professor Raimo P. Hämäläinen

Instructors: Professor Ahti Salo, M. Sc. (Tech.) Seppo Noppari

Abstract:

The Quality of Service has become an essential matter now that Internet services have come into general use. Even temporary failures may cause user dissatisfaction, resulting in customer loss for the Internet Service Provider. Providing uninterrupted service sets high requirements for service production systems. There is also a need for methods for risk assessment of such systems.

The design of a networked computing system, which provides Internet services, can be created according to several recommendations and standards. In addition, there are several frameworks and methods for risk assessment. Several reference standards are introduced in this thesis to provide foundation for system design and analysis.

In this thesis, I have used a dependability framework to collect the most important properties (availability, reliability, maintainability and security) of Internet service production systems for my analysis. To provide a generic view, I have created a few simple structural models first. Fault trees and reliability block diagrams provide additional means of considering redundancy options among other design choices.

The dependability framework and various methods are studied by applying the SQUALE dependability construction process to the generic models of Internet service production systems. The hazard frequency and severity classes of the ALARP principle have been adjusted for the analysis of these systems.

The thesis includes a case study, which presents how one of the dependability attributes, availability, can be monitored using a polling system. However, polling data may provide misleading results for several reasons. Therefore, problems with polling data are also introduced. The ALARP classification is used to rate the hazards of the production system.

The results of this thesis can mainly be used in production planning and system design.

Keywords: modeling, dependability, risk assessment, Internet services

Language: English