This is the PCN Linux Cluster, "amdahl".


Named after Gene Amdahl, the creator of Amdahl's Law, which governs the speedup of using parallel processors on a problem, versus using only one serial processor.

More information on Amdahl's law can be found here.

The "amdahl" cluster consists of

  • 12 dual Athlon MP 1800 compute nodes,
  • three 4-cpu 2 GHz Xeon nodes
for a grand total of 36 processors, each with at least one GB of RAM. All nodes are connected by gigabit Ethernet.

2 TB of storage are devoted to the cluster, and access to every other storage resource (over two more terabytes) in the department is available from within the cluster as well.

Resources in this cluster are accessible to the Open Science Grid, as the Purdue-Physics Compute Element.

Introductory documents on the Physics Department's Linux Cluster are available now. [Postscript]  [PDF]

New to the Physics Department Cluster?
Want to get a computation running quickly?
Read the Quick Start guide. [Postscript]  [PDF] 

The Cluster User Guide is available: [Postscript]  [PDF]  [HTML]

May 29, 2002: Full benchmark information is now available: [Postscript]  [PDF]

Check the cluster status with the Ganglia cluster toolkit.
Check the contents of the PBS Queue.

Job running on the cluster should not be done interactively, with the exception of the interactive node, 'node1'. Node1 should be used for compiling, and short-run tests. Actual execution of jobs should be submitted with PBS, with the qsub command. qsub uses a script file to inform PBS about the job being submitted. An example PBS script file is provided, with comments illustrating useful options. Additionally, an example of running an MPICH parallel program from within PBS can be found as well.

Further reference on PBS commands can be found here:

Other documentation resources


Purdue Homepage Purdue Search Purdue Maps Purdue Directories Copyright © 2003, Purdue University, all rights reserved. An equal access/equal opportunity university.
Purdue Disclaimer Nondiscrimination Policy