Search citation statements
Paper Sections
Citation Types
Year Published
Publication Types
Relationship
Authors
Journals
In this paper we present some results obtained from using simulation as a tool for evaluating alternative methods of improving system performance in network computers. We begin by describing a GPSS model which was developed to evaluate the current computer center operations at the University of Illinois. While using actual data from the center to verify that the rriodel accurately predicts each job's total time in the system, we evaluate the resource utilization within the center. Then we use the simulator to show that a proposed dynamic priority assignment algorithm yields better throughput than the existing algorithm while maintaining a higher level of CPU and memory utilization. Next, turning our attention to a hypothetical network of three centers, we use a generalization of our model to demonstrate the effects of load leveling between tenters on the total system throughput. Here we also explore the merits of "-pay-for-priority" schemes in both a single center and in a network of three centers.
In this paper we present some results obtained from using simulation as a tool for evaluating alternative methods of improving system performance in network computers. We begin by describing a GPSS model which was developed to evaluate the current computer center operations at the University of Illinois. While using actual data from the center to verify that the rriodel accurately predicts each job's total time in the system, we evaluate the resource utilization within the center. Then we use the simulator to show that a proposed dynamic priority assignment algorithm yields better throughput than the existing algorithm while maintaining a higher level of CPU and memory utilization. Next, turning our attention to a hypothetical network of three centers, we use a generalization of our model to demonstrate the effects of load leveling between tenters on the total system throughput. Here we also explore the merits of "-pay-for-priority" schemes in both a single center and in a network of three centers.
A number of performance prediction methods are available to IBM marketing personnel. This paper describes one such method, which predicts the effects ofchanges in IBM 3790 and 8100 distributed processing systems and in teleprocessing networks. Such changes may involve system features (such as line protocols), the introduction of new applications, or volume growth in an otherwise static system. The technique makes use ofa detailed simulator, informally called FIVE, in conjunction with a system monitor and data analysis program. Its use can make substantial performance information available at relatively low cost.From a capacity planning viewpoint, a system can be divided into three broad areas: host processor, communication network, and distributed intelligence, as in the IBM 3790 and 8100 systems.i" Attention in the past has generally been focused on the host system, but the sharply declining cost of central hardware, combined with the rapid growth of teleprocessing and distributed processing, has increased the importance of communications and distributed intelligence.Systems with dozens of distributed processors are no longer uncommon. Managing the growth in capacity of such systems requires close attention to key performance considerations. The basic requirement for capacity planning is predicting how changes will affect system performance, thereby permitting the best course of action to be followed. For purposes of the following discussion, the key performance elements are assumed to be response time (as perceived by a terminal operator) and throughput (volume of work per unit time, such as characters printed per second). Methods for predicting system performance include "educated guesswork," benchmarking, and the use of computer models (either analytic or simulative, as discussed below). The first method Copyright 1980 by International Business Machines Corporation. Copying is permitted without payment of royalty provided that (I) each reproduction is done without alteration and (2) the Journal reference and IBM copyright notice are included on the first page. The title and abstract may be used without further permission in computer-based and other information-service systems. Permission to republish other excerpts should be obtained from the Editor.is time-honored and probably the most widely used. It works best when the system is comfortably overdesigned, allowing a safe margin for error. This casual approach to performance prediction has become increasingly unsatisfactory because of the growth of on-line systems (with their stringent performance requirements) and diminishing tolerance for gross overdesign.Benchmarking (the execution of actual applications on the proposed system configuration) is expensive and inflexible, since it is difficult to examine a wide range of configuration alternatives. For a combination of flexibility (in specifying both system configuration and applications at any level of detail) and reasonable cost, computer models have come into widespread use. Such models are programs that ...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.