Computer System Simulation: An Introduction
ACM Computing Surveys (CSUR)
Trace driven modeling and analysis of CPU scheduling in a multi-programming system
Proceedings of the SIGOPS workshop on System performance evaluation
A Petri Net model of the CDC 6400
Proceedings of the SIGOPS workshop on System performance evaluation
Design of a Computer—The Control Data 6600
Design of a Computer—The Control Data 6600
Simulation of an ECS-based operating system
AFIPS '67 (Spring) Proceedings of the April 18-20, 1967, spring joint computer conference
Who are the users?: an analysis of computer use in a university computer center
AFIPS '71 (Spring) Proceedings of the May 18-20, 1971, spring joint computer conference
Trace driven analysis of a batch processing system
ANSS '73 Proceedings of the 1st symposium on Simulation of computer systems
Some steps toward an information system performance theory
ACM SIGMETRICS Performance Evaluation Review
Evaluation nets for computer system performance analysis
AFIPS '72 (Fall, part I) Proceedings of the December 5-7, 1972, fall joint computer conference, part I
Objectives and problems in simulating computers
AFIPS '72 (Fall, part I) Proceedings of the December 5-7, 1972, fall joint computer conference, part I
| Hi-index | 0.00 |
A computer center typically faces questions of how to deal with a growing load in the face of tight financial constraints and with the need for lead time for planning ways to meet the demand. The needs may be met by altering the equipment configuration, changing priority algorithms and other features of the operating system, controlling the time scheduling of various classes of job load, or perhaps shifting load from one machine to another if the center is large enough to have such a capability. It is often difficult to gather enough data and insight to show the direction these changes should go and to support the decision to do so. One useful set of tools is provided by simulation backed up by measurements required for validation. However, it is all too common to find that fear of interruptions of the computing center's service to user, combined with an overworked systems programming staff, prevents insertion of the desired measurement probes into the operating system. Then one is restricted to measures that can be derived from the normal accounting log, or "Dayfile", and to software probes that can be injected in the guise of user's programs. In spite of these limitations, useful results can be obtained, and this paper describes validation of a simulation model of a multiprogramming system, the Control Data 6400, making use of these restricted measurements.