Studies of an Optical Multi-Processor Interconnect

Abstract

Optical fiber provide a very high capacity medium for point to point transmission of data. It is not straight forward to achieve the same level of success in networked fiber communication systems. A carefully designed system can reach terabyte-per-second capability. The purpose of this study was to investigate the performance of a generic optical distributed interconnect architecture for parallel processing. In particular, the target was the general purpose, easily programmable tightly coupled multiple-instruction stream, multiple data stream (MIMD) systems. We have proposed a fiber-optic interconnect system that is aimed at exploiting the advantages provided by both optics and electronics to provide a low latency, high throughput communication between processors and memory modules in a shared memory system. Decisions taken during any new design are to be based on the basic requirements placed on the system, the strengths and weakness of the implementation domain as well as the identification of the problems that exist in past designs.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1994
Accession Number
ADA277093

Entities

People

  • Jon R. Sauer

Organizations

  • University of Colorado Boulder

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Application Software
  • Communication Systems
  • Computer Programming
  • Computers
  • Computing System Architectures
  • Delay Lines
  • Energy Bands
  • Hot Spots
  • Multiple Access
  • Multiplexing
  • Nets
  • Network Topology
  • Optical Interconnects
  • Plasma Opening Switches
  • Simulators
  • Throughput
  • Two Dimensional

Readers

  • Optical Physics and Photonics.
  • Parallel and Distributed Computing.

Technology Areas

  • Microelectronics