Date of Award

Fall 1992

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Feng, Xin

Second Advisor

Allgaier, Glen

Third Advisor

Niederjohn, Russell

Abstract

A new combined neural network and data visualization approach is proposed and applied to the capacity planning of computer resources based on a centralized computer system. A back-propagation neural network is employed to model the current system configuration, including memory, central processing unit (CPU) and input/output (I/0) resources. Response predictions are made by estimating future workload demands. The prediction of system response is necessary to determine the lifetime of current computer equipment and also to guarantee performance requirements defined by the organization as service levels. The study demonstrates that the new proposed method provides a better approach for system planning and configuration than traditional methods. Moreover, the process by which the neural network has been uniquely adapted to computer performance prediction makes this new approach convenient for a wide range of general purpose applications of this type. The results are further enhanced by the application of data visualization technology. Visualizing neural network response surfaces provides insights into the operation of the network, the significance of various input classes, and the interpretation of the output response.

Share

COinS

Restricted Access Item

Having trouble?