Sidebar image

HTCondor Cluster at Math

Computing cluster is best suited for tasks consisting of a large number of similar jobs, which can be run in parallel. For example, same algorithm run with varying input data.

Math HTCondor cluster size changes over time, but usually has 40–50 processor cores available. Status (size and load) of the cluster can always be checked here.

Jobs are submitted using HTCondor software. HTCondor is free and open source, with binaries available for Linux and Windows. Installation instructions tailored for Math cluster are here: Linux, Windows. HTCondor is already installed and running on Math Dept. application servers gauss and doob as well as on Grad Lab Linux workstations.

If you have a fairly modern Linux workstation, you should consider installing HTCondor and adding it to Math cluster as an execute node. Jobs will only be scheduled to run on your machine when it was idle for a certain amount of time, which is configurable. Installation is very similar to that of a submit-only node, and the differences are pointed out in instructions for Linux above.

With HTCondor installed and running, jobs can be sent to the cluster by preparing so-called “submit” file and then issuing command:

   condor_submit submitfile
Here is a zip-file with some very basic examples of submit files, including those for Maple, Mathematica and Matlab. Please browse through HTCondor User's Manual to get an idea of what else is possible.