skip to main content
DOE Patents title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: High-performance parallel processors based on star-coupled wavelength division multiplexing optical interconnects

Abstract

As the performance of individual elements within parallel processing systems increases, increased communication capability between distributed processor and memory elements is required. There is great interest in using fiber optics to improve interconnect communication beyond that attainable using electronic technology. Several groups have considered WDM, star-coupled optical interconnects. The invention uses a fiber optic transceiver to provide low latency, high bandwidth channels for such interconnects using a robust multimode fiber technology. Instruction-level simulation is used to quantify the bandwidth, latency, and concurrency required for such interconnects to scale to 256 nodes, each operating at 1 GFLOPS performance. Performance scales have been shown to .apprxeq.100 GFLOPS for scientific application kernels using a small number of wavelengths (8 to 32), only one wavelength received per node, and achievable optoelectronic bandwidth and latency.

Inventors:
 [1];  [2];  [3]
  1. (Pleasanton, CA)
  2. (Castro Valley, CA)
  3. (Arvada, CO)
Issue Date:
Research Org.:
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
OSTI Identifier:
874546
Patent Number(s):
6411418
Assignee:
The Regents of the University of California (Oakland, CA) LLNL
DOE Contract Number:  
W-7405-ENG-48
Resource Type:
Patent
Country of Publication:
United States
Language:
English
Subject:
high-performance; parallel; processors; based; star-coupled; wavelength; division; multiplexing; optical; interconnects; performance; individual; elements; processing; systems; increases; increased; communication; capability; distributed; processor; memory; required; fiber; optics; improve; interconnect; attainable; electronic; technology; considered; wdm; optic; transceiver; provide; latency; bandwidth; channels; robust; multimode; instruction-level; simulation; quantify; concurrency; scale; 256; nodes; operating; gflops; scales; shown; apprxeq100; scientific; application; kernels; wavelengths; 32; received; node; achievable; optoelectronic; fiber optic; memory element; individual elements; parallel processor; processing systems; /359/

Citation Formats

Deri, Robert J., DeGroot, Anthony J., and Haigh, Ronald E. High-performance parallel processors based on star-coupled wavelength division multiplexing optical interconnects. United States: N. p., 2002. Web.
Deri, Robert J., DeGroot, Anthony J., & Haigh, Ronald E. High-performance parallel processors based on star-coupled wavelength division multiplexing optical interconnects. United States.
Deri, Robert J., DeGroot, Anthony J., and Haigh, Ronald E. Tue . "High-performance parallel processors based on star-coupled wavelength division multiplexing optical interconnects". United States. https://www.osti.gov/servlets/purl/874546.
@article{osti_874546,
title = {High-performance parallel processors based on star-coupled wavelength division multiplexing optical interconnects},
author = {Deri, Robert J. and DeGroot, Anthony J. and Haigh, Ronald E.},
abstractNote = {As the performance of individual elements within parallel processing systems increases, increased communication capability between distributed processor and memory elements is required. There is great interest in using fiber optics to improve interconnect communication beyond that attainable using electronic technology. Several groups have considered WDM, star-coupled optical interconnects. The invention uses a fiber optic transceiver to provide low latency, high bandwidth channels for such interconnects using a robust multimode fiber technology. Instruction-level simulation is used to quantify the bandwidth, latency, and concurrency required for such interconnects to scale to 256 nodes, each operating at 1 GFLOPS performance. Performance scales have been shown to .apprxeq.100 GFLOPS for scientific application kernels using a small number of wavelengths (8 to 32), only one wavelength received per node, and achievable optoelectronic bandwidth and latency.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2002},
month = {1}
}

Patent:

Save / Share: