Browse Prior Art Database

SPECIAL FEATURE: Some Considerations in the Design of Mainframe Processors with Microprocessor Technology

IP.com Disclosure Number: IPCOM000131437D
Original Publication Date: 1979-Aug-01
Included in the Prior Art Database: 2005-Nov-11
Document File: 9 page(s) / 37K

Publishing Venue

Software Patent Institute

Related People

Gary Tjaden: AUTHOR [+4]

Abstract

[Figure containing following caption omitted: Implementing mainframes with multimicroprocessors entails an extra cost for main memory. This often-overlooked phenomenon is analyzed here for several mainframe systems.] LSI technology, in the form of random-access memory chips and microprocessors, has led to the availability of computing systems priced very much lower than ";mainframe"; systems such as the Sperry Univac 1100/40 or the IBM 370/155. Mainframe systems, however, offer some features -- higher performance and fault tolerance, for example -- that ";microprocessor"; systems do not. Achieving mainframe functionality with LSI technology is of great interest to computer designers because of the anticipated dramatic improvement in cost-performance. One design approach that is often proposed is to interconnect a set of microprocessors so that each microprocessor executes different user jobs. The total throughput of this concurrently executing set of microprocessors is expected to be equivalent to the throughput of a single mainframe processor. This paper discusses some considerations which are often overlooked by proposers of this ";multimicroprocessor"; design approach. Performance in mainframe systems is often expressed in terms of millions of instruction executed per second. MIPS is usually determined by measuring the execution time of a representative mix of user jobs running on a system configuration in which main memory size and I/O capability (number and types of channels and peripheral devices) is well matched to central processor speed and user job requirements. The job mix will typically contain a large percentage of Fortran and Cobol programs, and a much smaller percentage of jobs written in several other languages such as Algal and PL/I. These jobs will be run under a specified operating system version, usually with some mix of both batch and interactive operation. The number of machine instructions executed and the time required for their execution are used to compute the performance in MIPS.

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

Page 1 of 9

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

This record contains textual material that is copyright ©; 1979 by the Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Contact the IEEE Computer Society http://www.computer.org/ (714-821-8380) for copies of the complete work that was the source of this textual material and for all use beyond that as a record from the SPI Database.

SPECIAL FEATURE: Some Considerations in the Design of Mainframe Processors with Microprocessor Technology

Gary Tjaden

Sperry Univac

Martin Cohn

Sperry Research Center

(Image Omitted: Implementing mainframes with multimicroprocessors entails an extra cost for main memory. This often-overlooked phenomenon is analyzed here for several mainframe systems.)

LSI technology, in the form of random-access memory chips and microprocessors, has led to the availability of computing systems priced very much lower than "mainframe" systems such as the Sperry Univac 1100/40 or the IBM 370/155. Mainframe systems, however, offer some features -- higher performance and fault tolerance, for example -- that "microprocessor" systems do not.

Achieving mainframe functionality with LSI technology is of great interest to computer designers because of the anticipated dramatic improvement in cost-performance. One design approach that is often proposed is to interconnect a set of microprocessors so that each microprocessor executes different user jobs. The total throughput of this concurrently executing set of microprocessors is expected to be equivalent to the throughput of a single mainframe processor. This paper discusses some considerations which are often overlooked by proposers of this "multimicroprocessor" design approach.

Performance in mainframe systems is often expressed in terms of millions of instruction executed per second. MIPS is usually determined by measuring the execution time of a representative mix of user jobs running on a system configuration in which main memory size and I/O capability (number and types of channels and peripheral devices) is well matched to central processor speed and user job requirements. The job mix will typically contain a large percentage of Fortran and Cobol programs, and a much smaller percentage of jobs written in several other languages such as Algal and PL/I. These jobs will be run under a specified operating system version, usually with some mix of both batch and interactive operation. The number of machine instructions executed and the time required for their execution are used to compute the performance in MIPS.

This method of determining computer system performance is oriented toward the user who is trying to compare the cost- effectiveness of computer systems on the basis of total system throughput as seen in his environment. The resultant performance measure depends on many parameters other than just the hardware technology with which the processor and main memory are implemented. These...