© 1981 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
Validation of an analytic model of computer performance

1 Department of Mathematics, University of Nottingham, University Park, Nottingham NG7 2RD, UK, 2 Department of Computer Science, ISSR, Cairo University, Cairo, Egypt
The objective of validating an analytic model for performance prediction of a computer system is to establish confidence in its use as a tool. In general the problem of validation has two main aspects. The first deals with the internal correctness of the model, in that its mathematical structure must satisfy the functions and logical sequences intended by the designer. This should be a basic objective of the analyst, and is the simpler of the two aspects. However, the more formidable aspect of validation lies in deciding how adequately the model represents actual systems and how far it reflects their major characteristics. The work in this paper deals with this latter aspect. A procedure for estimating input parameter values and output performance measures from available performance statistics and measurements is described. The model used is a class of queueing networks known as central-server, although the same procedure could be applied to other types of model.
Received May 1980.
* Department of Mathematics, University of Nottingham, University Park, Nottingham NG7 2RD, UK.
Department of Computer Science, ISSR, Cairo University, Egypt.