TAFSM

Team for Advanced Flow Simulation and Modeling



Home

Research Overview

Research Highlights

Projects

Takizawa Lab

Undergraduate Research

TAFSM Featured

TAFSM Recognized

Publication, Preprints

Currrent Team Members

Collaborators, Ex-Members

AHPCRC, History

Links

Next FSI short course

For more information:
  tezduyar@gmail.com

AHPCRC Bulletin: Summer 1997 - Volume 7 Number 3

New Computers for AHPCRC

Paul Muzio (AHPCRC-NetworkCS)

Tayfun Tezduyar and Congressman Martin Sabo at a press conference on 16 October 1997 announcing the award of funds to the AHPCRC for the procurement of new systems.
The AHPCRC has announced that it will acquire and install new HPC systems in early 1998. A contract has been signed with Silicon Graphics/Cray Research to acquire a CRAY T3E-1200 supercomputer, which will be located in Minneapolis. Also, negotiations are underway to acquire four IBM RS/6000 SP systems, which will be located in January and February 1998 at AHPCRC partner sites-Clark Atlanta, Florida A&M, Howard, and Jackson State Universities. Funding for the systems was made available in the FY 1997 Defense Appropriations Act through the efforts of Representative Martin Sabo of Minnesota.

The AHPCRC will receive delivery of the first (serial number 1) CRAY T3E-1200, in January 1998. The T3E-1200 is similar in appearance to the SGI/CRAY T3E-900, but uses faster processors and memory than the T3E-900. The AHPCRC's T3E-1200 will have 256 user and 16 system processor elements, with a total peak performance of 326 GigaFLOPS and a total of 139 Gigabytes of distributed memory. At 139 Gigabytes, this T3E is one of the largest computers in the world in terms of total memory. The system will also have 500 Gigabytes of disk storage and 1 Terabyte of HIPPI RAID storage. It is expected that the four IBM RS/6000 SP systems will have a total of approximately 80 nodes with over 40 Gigabytes of distributed memory. All five systems are scalable, distributed memory systems and will be used to support the Center's research on critical Army technology issues in the fields of 3D simulation and modeling, computational structural mechanics, environmental sciences, advanced materials science, and enabling technologies, and will be interconnected by a high speed communications network. The acquisition, installation and operation of the systems is being managed by the AHPCRC's system integrator, Network Computing Services, Inc.

Walter Sturek, Paul Muzio, and Gary Hansen viewing the AHPCRC's T3E-1200 being built at Cray Research's facilities in Chippewa, Wisconsin.
According to Tayfun Tezduyar, Distinguished McKnight University Professor and AHPCRC Director, "Researchers at our Center address technological issues as varied and complex as developing methods to control pollution and designing safe deployment of parachute systems used in the crew return vehicle of the international space station. This new computational power will enable us to build a flexible simulation and modeling environment to address these pressing technological issues."

Tezduyar went on to say "These new systems, with their high computational power, will be important in maintaining our internationally recognized lead in HPC research and education, ... and will be a powerful tool for our work in developing the computational methodology for 3D modeling and simulation of flow around, and the dynamics of, complex structures such as ram-air parachutes. These parachutes are planned to be used for the delivery of large quantities of material supplies and in support of the crew return vehicle for the space station. These systems will also be a critical asset in our ability to develop improved computational methods for groundwater modeling, drug design, and materials science."

The AHPCRC's CM-5 has had a long, productive life, having been in production for over 6 years.
The CRAY T3E-1200 and IBM RS/6000 SPs will replace the Center's Thinking Machines CM-5, which is scheduled to be turned off on 28 February 1998. This system was the first CM-5 and was installed in August 1991. In its current configuration, it has a total capability of 115 GigaFLOPS with 29 Gigabytes of memory. It was a pioneering system in massively parallel computing and remains as one of the largest and most heavily used scalable systems in operation in the DoD inventory. A comparison between the T3E-1200 and CM-5 is provided in Table 1.

  CRAY T3E-1200 TMC CM-5
Peak System Performance 326 GigaFLOPS 115 GigaFLOPS
System Memory 139 Gigabytes 29 Gigabytes
Number of Processors 272 (includes 16 system processors) 896
Peak Processor Element Performance 1200 MegaFLOPS 128 MegaFLOPS
Memory per Processor 512 Megabytes 32 Megabytes
Processor Type DEC Alpha EV 5.6 SPARC-2 with custom vector processors
System Disk 500 Gigabytes 125 Megabytes
HIPPI Disk 1 Terabyte 288 Gigabytes
Primary Coolant Liquid Air
Programming Models Message passing, HPF Data parallel
Table 1. Comparison between the T3E-1200 and the CM-5.