Linux clusters are a cost-effective and versatile way to get your jobs run.

See also:

• NamdOnScyld
• NamdOnClustermatic
• NamdOnRocks
• NamdOnPBS
• NamdOnGridEngine
• NamdOnMPICH
• NamdOnLAMMPI
• NamdOnMyrinet
• NamdOnElan

A good mailing list for life science oriented clusters is the bioclusters mailing list. Archives and subscription information can be found online at https://bioinformatics.org/mailman/listinfo/bioclusters

A good book on running a Linux machine is Running Linux, http://www.oreilly.com/catalog/runux4/

A good book on building a Linux cluster is the online and under-construction Engineering a Beowulf-Style Compute Cluster by Robert G. Brown, http://www.phy.duke.edu/~rgb/Beowulf/beowulf_book.php

A free and reasonably simple to install queueing system is Sun's Grid Engine, http://gridengine.sunsource.net/ (see NamdOnGridEngine)

Useful testing tools are cpuburn http://pages.sbcglobal.net/redelm/ and netperf http://www.netperf.org/netperf/NetperfPage.html

There are several cluster-oriented distributions of Linux:

Scyld (see NamdOnScyld) and Clustermatic (see NamdOnClustermatic) use Bproc (http://bproc.sourceforge.net/) to provide a unified process space and eliminate installing Linux on the slaves completely. These are good for running a single NAMD job on the entire cluster and in some other circumstances, but programs expecting a full Linux environment will be disappointed.

NPACI Rocks (http://rocks.npaci.edu/) maintains node consistency by using RPM to completely reinstall the slaves to propagate any minor change.

OSCAR (http://oscar.sourceforge.net/) takes a less radical approach, bundling tools for configuring and maintaining a traditional cluster.