detailed information about saguaro nodes 9-30-13

2014-08-29

++ detailed information about saguaro nodes 9-30-13

The cluster is a mixture of Intel architectures from 2009 to the present. Some are faster than others, and the speed also depends on the type of work your programs are doing.
Some are using DDR Infiniband, some QDR, and some FDR.

Here is a list of the current computers in the cluster:

# Node names and architectures:
# * s61 sandy bridge EP; 16 cores/node; 16 nodes/chassis; 64GB RAM/node; FDR IB (PowerEdgeM620)
# * s23 - s42: harpertown; 8 cores/node; 16 nodes/chassis; 16GB RAM/node; DDR IB (PowerEdgeM600)
# * s54 - s57: nehalem; 8 cores/node; 16 nodes/chassis; 24GB RAM/node; DDR IB (PowerEdgeM610)
# * s58 - s59: westmere X5650; 12 cores/node; 16 nodes/chassis; 48GB RAM/node; QDR IB (PowerEdgeM610)
# * s63: westmere X5670; 12 cores/node; 16 nodes/chassis; 96GB RAM/node; QDR IB PowerEdge710HD)
# * fn1 - fn12: tigerton; 16 cores/node; 1 node/chassis; 64GB RAM/node; DDR IB(PowerEdgeR900)
# * s60 : X7560 32 cores/node; 1 node/chassis; 1TB RAM/node; X7560 QDR IB (PowerEdgeR910)

If you want to run your programs always on the same computers, add a FEATURE to your qsub script. such as:

PBS -l nodes=32:harpertown

The time that it takes to complete a job also depends on just how busy the computer(s) are where your jobs are running, whether there are any other jobs running on that particular computer, how busy the Infiniband network is, how busy the ethernet network is, how much disk I/O is being performed, and so on.

Let us know how we can help.