Browse Prior Art Database

A Dual Priority MVA Model for a Large Distributed System: LOCUS

IP.com Disclosure Number: IPCOM000128337D
Original Publication Date: 1984-Dec-31
Included in the Prior Art Database: 2005-Sep-15
Document File: 13 page(s) / 43K

Publishing Venue

Software Patent Institute

Related People

Joseph Betser: AUTHOR [+5]

Abstract

The analytic issues of distributed systems are rich and complex. Moreover, as the builders of such systems are becoming increasingly ambitious, even larger systems are being conceived. This upscaling in the system dimensions leads to an explosion in the combinatorial space of design issues involved. It is becoming more apparent than ever that a thorough understanding of the theoretical foundation of these systems should be sought, in order to generate cost-effective designs.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 10% of the total text.

Page 1 of 13

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

A Dual Priority MVA Model for a Large Distributed System: LOCUS

Joseph Betser, Mario Gerla, and Gerald Popek

Computer Science Department University of California. Los Angeles Los Angeles, CA 9004

May 1984 * This research has been supported by the Advanced Research Projects Agency, under contract DSS-NIDA-903-8'1.-C-0189. A Dual Priority MVA Model for a Large Distributed Systems LOCUS ~

A twofold extension to an MVA Eased analytical model is presented. A dual priority approximation is developed for a mixed queueing network. Improved fit to experimental results, obtained from our distributed system, LOCUS, is observed. Issues of file replication in large (up to 100 sites) distributed systems are examined, with respect to performance measures. Important design guidelines are obtained. Finally, the computational re-quirements of our model are assessed. Performance prediction of yet larger systems appears quite feasible.

1. Introduction

The analytic issues of distributed systems are rich and complex. Moreover, as the builders of such systems are becoming increasingly ambitious, even larger systems are being conceived. This upscaling in the system dimensions leads to an explosion in the combinatorial space of design issues involved. It is becoming more apparent than ever that a thorough understanding of the theoretical foundation of these systems should be sought, in order to generate cost- effective designs. This paper addresses one such design issue, namely the problem of supporting replicated files* in the context of system scaling. Replicated files increase the availability and reliability of data in distribut-ed systems. However, this feature has a cost in terms of system resources. We study these resource demands, and provide a general analytical model, which is then applied to a particular real system for validation and further performance prediction.

1.1 The LOCUS System

LOCUS, the test system used for experimental comparison with the analytic model predictions, is a distributed operating system which supports transparent access to data through a network wide system, permits automatic replication of storage, supports transparent distributed process execution, such as nest-ed transactions, and is upward compatible with Unix. Partitioned operation of subnets and their dynamic merge are also supported (POPE811, (WALIC83b]. LOCUS is functioning as the distributed operating sys-tem of the UCLA Center of Experimental Computer Science (CECS) network. It has been operating at UCLA since 1981, and is currently run on 17 VAX 11/750s, connected by a 10 Mbit/sec Ethernet, as described in Fig 1.

1.2 High Priority Servers

In LOCUS, application program access to remote and local resources is achieved using the same system semantics. This is the notion of Network Transparency. If performance in

UCLA Page 1 Dec 31, 1984

Page 2 of 13

A Dual Priority MVA Model for a Large Distributed...