| |
|
|
|
|
 |
NCSA NEWS |
|
|
 |
 |
|
In January, NCSA and IBM announced plans to install two Linux clusters, creating the world's fastest Linux supercomputers in academia (see http://access.ncsa.uiuc.edu/Releases/010116.IBM.html). Delivery of the first cluster, consisting of IBM eServer x330 thin servers and based on Intel's Pentium III architecture, began Feb. 5. But before any computing equipment was unpacked, more than a year's worth of work went on behind the scenes among technical teams at NCSA, IBM, Intel, and Myricom. Below is a timeline that documents this collaborative effort between the private and public sectors in an effort to create a two teraflop Linux cluster computing environment.
 |
| The first shipment of IBM eServer x330 thin servers awaits shipment to NCSA |
 |
 |
| Setting up the racks that will hold NCSA's first terascale cluster |
September 1999
NCSA begins working with Intel on porting codes to the 64-bit Itanium processor.
November 1999
NCSA, Intel, and SGI demonstrate the first computation cluster using the Itanium processor at SC99 in Portland, OR. A cluster of four SGI Itanium machines running Linux executes the Cactus code using MPI for message passing between four nodes. The demonstration runs live before hundreds of conference attendees.
May 2000
NCSA works with Myricom and Intel to create a prototype cluster of four four-processor Itanium servers. Cactus, MILC, and sPPM are among the first scientific codes to be tested.
Summer 2000
Performance evaluation on the 16-processor cluster continues. NCSA's cluster team works with the Alliance application teams and engineers from Intel on performance for three scientific codes using early Itanium compilers. Also, Myricom engineers work with NCSA on making the Myrinet interconnect fully functional with Itanium systems running Linux.
October 2000
NCSA's first presentation on an Itanium cluster using Myrinet takes place at the Intel eXHANGE in San Francisco. Cactus and sPPM are demonstrated, including interactive steering of the Cactus computation.
 |
| Hydrodynamics simulation by the University of Minnesota's Paul Woodward created on a prototype Alliance Itanium cluster using sPPM. |
November 2000
At SC2000 NCSA officially announces performance levels of 650 megaflops for sPPM, MILC, and Cactus on a single processor prototype Itanium-based server running 64-bit Linux. In addition, NCSA announces that Alex Granovsky of Moscow State University achieved exceptional performance on a prototype Itanium cluster running Microsoft's 64-bit Windows. Granovsky ran a version of the General Atomic and Molecular Electronic Structure System (GAMESS) ab initio quantum chemistry program on an eight-processor Itanium-based server cluster at NCSA. Performance on an eight-processor Itanium cluster was more than 12 gigaflops.
December 2000
32 dual-processor Itanium-based prototype server machines running Turbolinux arrive at NCSA from IBM.
January 2001
IBM and NCSA jointly announce plans to install two IBM Linux clusters at NCSA. The clusters' combined computing power of two teraflops will make them the world's fastest Linux supercomputers in academia.
February, 2001
Itanium processor performance on the MILC code exceeding 1 gigaflop per second is announced at the LinuxWorld Conference in New York.
IBM Global Services installs the first Linux cluster at NCSA. It consists of 512 IBM eServer x330 thin servers, each with two 1 GHz Intel Pentium III processors running Red Hat Linux.
![[Up]](feat_up.gif)
Access Online | Posted 2-13-2001
|
|
|
|
|
|
