CEC Theses and Dissertations

Title

A New Design Approach to an Expert System Network Manager Based on an Analysis of Contemporary Distributed Networks

Date of Award

1996

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Graduate School of Computer and Information Sciences

Advisor

S. Rollins Guild

Committee Member

Michael Moody

Committee Member

Laurie Dringus

Abstract

Today's organizations need optimized computing environments that integrate diverse hardware, operating systems and applications, sometimes at numerous locations. The rapid growth in size and complexity of distributed networks and the use of heterogeneous components require effective tools to manage these resources. The system management aspect of distributed networks is one of the least mature components of the environment, and unfortunately it is receiving little attention by systems developers for various reasons.

A new model of an expert network system network manager has been designed based on an analysis and evaluation of major contemporary distributed network management systems: (i) DEC PolyCenter network management system, (ii) Hewlett-Packard Open View network management system, (iii) IDM NetView network management system, and (iv) Sun Microsystems SunNet network management system. These network management systems can handle network operations, administration, maintenance and provisioning functions with human intervention, but they cannot handle or predict unforeseen network fault situations; autonomous capabilities are entirely missing. A model of an expert network management system that utilizes a rule-based paradigm for decision making and neural networks for fault-recognition and -prediction has been proposed. The design includes all the functionalities recommended by International Standard Organization (ISO) for Open Systems Interconnection (OSI) model. The proposed model also possesses existing operational features of contemporary distributed network management systems, and is capable of providing corrective network maintenance actions manually as well as automatically. A prototype was built and tested to authenticate the new model. Test results indicate that the combination of a rule-based paradigm with neural network fault-recognition and -prediction capabilities can be useful in the design of successful future distributed network management systems.

This document is currently not available here.

  Link to NovaCat

Share

COinS