ASCI Red
From TOP500 Supercomputing Sites - Wiki
[edit] Description
ASCI Red or ASCI Option Red, is a supercomputer installed at Sandia National Laboratories, located in Albuquerque, New Mexico. On June 29, 2006, Sandia National Laboratories announced that ASCI Red is decommissioned.
The project was a collaboration between Intel corporation and Sandia Labs, as part of the U.S. Government's Accelerated Strategic Computing Initiative (ASCI).
It was built as stage one of the Accelerated Strategic Computing Initiative (ASCI) by the United States Department of Energy and the National Nuclear Security Administration to build a simulator to replace live nuclear weapon detonation following the moratorium on underground testing. ASCI Red was retired from service in September, 2005 at the conclusion of government fiscal year 2005.
It is a mesh-based (38 X 32 X 2) MIMD massively-parallel processing machine consisting of 4510 compute nodes, 1212 gigabyte of total distributed memory and 12.5 terabyte of disk storage. The original incarnation of this machine used Intel Pentium Pro processors, each clocked at 200 MHz. These were later upgraded to Pentium II OverDrive processors. The current system has a total of 9298 Pentium II OverDrive processors, each clocked at 333 MHz. The system consists of 104 cabinets, taking up about 2500 square feet (230 m²). The system was designed to use commodity mass-market components and to be very scalable.
The original ASCI Red is notable for being the first computer on Earth to bench above 1 TeraFLOPS on the Linpack benchmark (1996), as noted in Top500 Supercomputer sites. After being upgraded with Pentium II Overdrive processors, the computer has demonstrated Linpack performance above 2 TeraFLOPS.
Different partitions of the machine use different operating systems. To the programmer, it looks like a normal Unix machine, running "Teraflops OS", Intel's distributed OSF/1 AD-based system originally developed for the Paragon XP/S supercomputer. The compute partition processors run Sandia's very light-weight "Cougar" operating system which traces its heritage back to the SUNMOS kernel developed for the compute nodes of the Paragon.
System has two "TeraFlop" LANs, the SRN TeraFlop LAN and the SCN TeraFlop LAN. The latter is for classified computing. Each TeraFlop LAN consists of a server, an end of "janus" network, and a large tape storage device. Janus has two ends, and a middle section. In its full configuration, janus consists of four rows, each with a restricted end and a secure end. The restricted ends of the rows appear to the users as a single machine, and the secure ends of the rows appear as a single machine. The center sections of the rows are switchable as a unit between the two ends.
[edit] Information Tables
| Vendor | Intel |
|---|---|
| System designer | Intel |
| Model name(s) | ASCI Red |
| Time of manufacture | 1997 |
| Number of systems sold | 1 |
| Resellers | - |
| GENERAL INFORMATION | |
|---|---|
| Class | MPP |
| Type | Distributed-memory multi-processor |
| Operating system | Cougar (Intel's port of Puma based on SUNMOS) for compute partition, Teraflops OS (Intel's distributed UNIX) for other partitions |
| Compilers | FORTRAN77, FORTRAN90, C, C++ |
| Other specific software | MPI |
| Affiliated technologies | - |
| CPU generation used | Pentium Pro (Pentium II OverDrive since 1999) |
| CPU model | 200MHz / 200Mflops Pentium Pro (333MHz / 333Mflops Pentium II OverDrive since 1999) |
| DRAM type | (Processor to Memory: 533 MB/s) |
| FPGA | n/a |
| System cooling | Air |
| Floor space | 2500 square feet or 230 m² (104 Cabinets: 76 Computer / 8 Network / 20 Disk) |
| NETWORK | |
| Node Interconnect | 800 MB/s (bi-directional) per Node |
|
Network switch / Size / Quantity | n/a |
| Network topology | Mesh (38 X 32 X 2) and indirect access |
| Management network | two TeraFlop LANs (Ethernet/ATM) |
| BUILDING BLOCKS | ||
|---|---|---|
| Block name --> | Compute Node | System |
| Structure | basic block | 4510 Compute Nodes (4736 CN since 1999) |
| Number of CPU | 2 | 9298 including I/O, Service nodes, etc. (9472 CPU since 1999) |
| Memory size (max) | 256MB | 1212 GB (including I/O nodes) |
| Other equipment | - | 12.5TB RAID Disk Storage (73 Disk I/O Nodes), 52 Service Nodes, 2 System Nodes |
| Power consumption | - | 850 kW |
| Theor.performance | 400 Mflops (666 Mflops since 1999) | 1830.4 Gflops (3154 Gflops since 1999) |
| BENCHMARKS | ||
|
HPL Benchmark: CPUs / Rmax / efficiency / Date | - | 9472x 333MHz / 2121GF / 67.2% / 1999 9152x 200MHz / 1338GF / 73% / 1997 |
| Application performance | - | - |
