The distributed computing cluster consists of six computers (also called "nodes"). One node is designated as the "head" or "master" node; this is the computer that uses interact with directly, and acts as the co-ordinator for all computing jobs submitted to the cluster. The other five nodes are called "compute" or "slave" nodes, and perform computations under the direction of the head node. The cluster is on its own private subnetwork. Only the head node is accessible from the outside world; the compute nodes are only accessible from the head node.

The cluster is built with Apple's Xserve hardware. It runs the Mac OS X operating system, which is a variant of UNIX (or more specifically, BSD). All nodes share a filesystem, which facilitates sharing information between nodes. The standard GCC compiler is available; this can compile programs in C, C++, and Fortran. Among high level programming languages, R is supported.

We have implemented distributed computing using the snow and Rmpi add-on packages to R. This method of distributing computations will work for a wide range of statistical applications including those that require writing specialized code in C or Fortran. Please see Steve Shiboski's tutorial for step-by-step instructions.