From CMMS
News
[edit] Monday, July 6, 2009
The Center for Simulation & Modeling has acquired a 800 core cluster from Penguin Computing. This new cluster has 100 nodes each with two quad-core Intel Nehalem CPUs. The computer nodes are connected via a high speed Infiniband network, which will deliver exceptional performance for parallel calculations using large numbers of CPUs. The new computer cluster will run the Scyld and TaskMaster software from Penguin Computing which will greatly facilitate management of the cluster and effective utilization of the computational resources.
The February 2008 Newsletter is now available.
[edit] Thursday June 5, 2008
Kickoff event for the University's new Simulation and Modeling Initiative will be October 24, 2008.
More information will be forthcoming.
[edit] Sunday June 1, 2008
kohn cluster is online.
The latest addition to CMMS, a 48 node cluster, with two quad-core Xeon CPU's per node and Infiniband interconnects is now online.
[edit] Thursday January 16, 2008
Titles and abstracts have been posted for CMMS Distinguished Lecturer Prof. Wilfred van Gunsteren's lectures on March 25 and 26, 2008.
[edit] Tuesday January 8, 2008
Gaussian03 Revision E.01 available on IBM p655 Machine
Release notes have been published on the Gaussian03 website.
The PBS script for this version is
/usr/local/bin/gog03E.01
The previous revision can be accessed using
/usr/local/bin/gog03
[edit] Wednesday January 2, 2008
Gaussian03 Revision E.01 available on Markov Cluster
Release notes have been published on the Gaussian03 website.
The PBS script for this version is
/usr/prog/bin/G03E.01pbs
The previous revision can be accessed using
/usr/prog/bin/G03pbs
[edit] Monday November 26, 2007
Parallel Computing Workshop at PSC
December 13 – 15
The Pittsburgh Supercomputing Center will host a second three-day workshop covering advanced topics in the Message Passing Interface. This workshop will build on topics covered in the July workshop.
The workshop is free but registration is required as space is limited. To register for the work shop, send an email to Tom Maiden by December 10, 2007.
[edit] Tuesday October 2, 2007
New policy regarding the verlet queue on the Markov Cluster, specificially that this queue is intended for parallel programs compiled to efficiently use the InfiniBand interconnect. Details are on the cluster page.
[edit] Monday October 1, 2007
The Torque PBS configuration on the Markov cluster has changed. Specifically, the node-classification has been replaced with explicit queues such that a job to be on small_io nodes is now submitted as:
#PBS -l nodes=1:ppn=1 #PBS -q small_io
For further information, see the cluster page and the Torque user guide.