Browse Prior Art Database

ON ENHANCING THE IDEALIZED CPU-I/O AND I/O-I/O OVERLAP MODELS THROUGH THE USE OF MARKOV PROCESSES

IP.com Disclosure Number: IPCOM000128446D
Original Publication Date: 1983-Apr-01
Included in the Prior Art Database: 2005-Sep-16
Document File: 11 page(s) / 42K

Publishing Venue

Software Patent Institute

Related People

Maletz, Mark C.: AUTHOR [+3]

Abstract

A Modeling technique is developed for studying processor overlap in time-shared and multiprogrammed computing systems, based on the Markov process methodology. Measures of resource utilization, response time, and system throughput are derived from the statistics of the Markov process equilibrium. Techniques for identifying potential system bottlenecks and for determining summary statistics about user transition behavior are also discussed. The result is an enhancement of traditional overlap models with greater generality and improved system characterization capabilities.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 11% of the total text.

Page 1 of 11

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

ON ENHANCING THE IDEALIZED CPU-I/O AND I/O-I/O OVERLAP MODELS THROUGH THE USE OF MARKOV PROCESSES

THE UNIVERSITY OF MICHIGAN COMPUTING RESEARCH LABORATORY1 Mark C. Maletz

CRL-TR-7-83

APRIL 1983

Room 1079, East Engineering Building
Ann Arbor, Michigan 48109
USA
Tel: (313) 763-8000

Abstract

A Modeling technique is developed for studying processor overlap in time-shared and multiprogrammed computing systems, based on the Markov process methodology. Measures of resource utilization, response time, and system throughput are derived from the statistics of the Markov process equilibrium. Techniques for identifying potential system bottlenecks and for determining summary statistics about user transition behavior are also discussed. The result is an enhancement of traditional overlap models with greater generality and improved system characterization capabilities.

[ Chapter ] 1. Introduction

As the complexity of modern time-shared and multiprogrammed computing systems continues to increase, the need for cost-effective analytical models of the performance of these systems also increases. System complexity can be broadly classified as belonging to one of four areas:

1. Changes in the number and speed of Central Processing Units (CPU's).
2. Changes in the number and nature of system peripherals (e.g., I/O devices).
3. Changes in the complexity of user behavior.
4. Changes in the amount of main storage available.

A computing system model is developed that is highly flexible with respect to the first three of the above areas. This model provides information that is of value when evaluating the performance of existing computing systems, selecting new systems, and identifying system resource bottlenecks.

The idealized CPU-I/O Overlap model was first proposed by Hellerman and Smith to analyze the expected throughput-of a computing system under certain "ideal" conditions.2 This computing system consisted of one central processing unit (CPU), one or two I/O channels, and main storage that could be partitioned into one, two, or three equally sized partitions. The "ideal"

1 Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the funding agencies.

2 Hellerman, H., and Smith, H.J. Jr. Throughput Analysis of Some Idealized Input, Output and Compute Overlap Configurations, Computing Surveys, 2,2 (June 1970), pp. 111 - 118.

University of Michigan Computing Research Laboratory Page 1 Apr 01, 1983

Page 2 of 11

ON ENHANCING THE IDEALIZED CPU-I/O AND I/O-I/O OVERLAP MODELS THROUGH THE USE OF MARKOV PROCESSES

conditions, aside from the particularly restrictive definition of the configuration of the computing system, were contained in a set of five assumptions:

1. Fixed record size.
2. For each input there must be an output.
3. Fixed number of records per block.
4. Fixed CPU compute time per re...