| Sequoia was acquired in summer 2008, and brought online for MCSR users as a 22-node, 176-core cluster in November 2008.
In June 2009, 8 more nodes were added.
In January 2010, 16 nodes were installed.
In July 2011, 38 nodes were built in.
In December 2011, 14 nodes were added.
In July 2012, 26 nodes were furnished.
These brought the system up to 124 nodes and 1304 CPUs.
Sequoia Cluster Diagram.
Sequoia Configuration
SGI Altix XE 250 Head Node:
The head node consists of a dual Intel Xeon Quad-core E5420 processors with 16 MB of memory on the node.
It also has 1333 MHz Front Side Bus, 12 MB L2 Cache and 250 GB Disk Drive.
Altix XE 310, Altix XE 320 and Rackable Compute Nodes:
Sequoia is a 124 node cluster with 22 Altix XE 310, 24 Altix XE 320, and 78 Rackable computing nodes.
Each Altix node consists of Dual Intel Xeon Quad-core E5420 Harpertown processors, a 1333 MHz Front Side Bus, a 12 MB L2 Cache, and a 250 GB Disk Drive.
Each Harpertown processor consists of 4 2.5 GHz cores (or, effectively, CPUs) and 8 GB memory.
42 Altix nodes are configured with 16 GB memory each (2GB per core), but 4 Altix fat nodes are configured with 32 GB memory each (4GB per core).
Each Rackable node has two six-core Intel Xeon X5650 (Westmere) processors, 36 GB of DDR3 RAM, a 2 TB internal SATA hard drive, and a QDR Infiniband network card.
Overall, Sequoia has 3.44TB memory.
In addition to the 250 GB Hard Drive available at each Altix node and 2 TB Hard Drive available at each Rackable node,
Sequoia has 14 TB of centralized scratch space (/tmp), 30 TB of permanent, recoverable user work space (/ptmp),
and 92 GB of centralized applications space (/usr/local/apps).
SGI Linux Compilers and Applications:
Compilers: Intel compilers for Linux (C/C++, Fortran) and GNU Compiler Collection (C/C++)
Libraries: SGI Message Passing Toolkit (MPT), Intel® Math Kernel Library (MKL) and FFTW
Applications: Gaussian 03 & 09, MolPro, CMPD and Amber 10.
| |
