User Tools

Site Tools


Arctic Region Supercomputing Center Pacman Cluster

What is Pacman?

Pacman is a 2816-processor Penguin Computing Cluster running Red Hat Enterprise Linux. With 12 Login Nodes, 256 Four Core Compute Nodes, 88 Sixteen Core Compute Nodes, 20 Twelve Core Compute Nodes, 4 Large Memory Nodes and 2 GPU nodes, Pacman offers our group the ability to conduct WRF-Chem model simulations at incredibly high resolution, within a reasonable amount of time.

What do we use Pacman for?

WRF-Chem is currently run on 16 and 4 core nodes and depending on the number of nodes requested, simulations can take only a few hours to run. I run on a single core, this could take days to complete. We are truly lucky to have such a great cluster, and great ARSC help desk, at this university.

In addition to the computational power of Pacman, we utilize the 89 TB Panasas version 12 and the 105 TB Lustre scratch file systems to store the large amounts of WRF-Chem output data needed to run simulations.

Output files are stored on the Sun SPARC Enterprise T5440 Server (named Bigdipper), which is directly accessible from Pacman and Fish shells. To make this storage system practically limitless in size, it is coupled to a StorageTek SL8500 Robotic Tape Library, capable of storing 29 petabytes of information. Wow!

pacman_cluster.txt · Last modified: 2012/09/13 13:41 by sean