This lesson guides you through the basics of submitting HTCondor jobs on the OSG Connect service, including a neuroscience-specific example of high-throughput computing from Chris Cox’s own research. HTCondor is software installed across the Open Science Grid (and many other computing systems, “pools”, or “clusters” around the world) that can schedule large numbers of computing “jobs” to run across large numbers of disparate, heterogeneous computers.
While some of the neuroscience-specific example leverages Matlab versions available through OSG Connect, the material about interacting with HTCondor would work on any HTCondor computing system.
Prerequisites
Prior experience with the Unix Shell (bash) is necessary, including skills covered in the morning portion of the tutorial, such as navigating a unix file system using absolute and relative paths, creating and editing text files on the command line, using wildcards, and creating shell scripts to automate the execution of several commands, in order. While there is some Matlab and Python code used in the neuroscience-specific example, you do not need to know how to program in these languages in order to submit the example jobs that use them.
00:00 | 1. Run Your First HTC Jobs | How can I use HTCondor to submit and track HTC work on the Open Science Grid? |
00:10 | 2. Reviewing HTCondor Job Files |
How do I know that my jobs completed successfully?
What files were created by my jobs and what information do they contain? |
00:20 | 3. Setting Up a Neuroscience HTC Workload | How can a real fMRI analysis be setup and launched? |
00:40 | 4. Setting Up a Neuroscience HTC Workload 2: setupJobs and stub.yaml | How does setupJobs turn a stub.yaml into a populated directory tree? |
01:00 | Finish |
The actual schedule may vary slightly depending on the topics and exercises chosen by the instructor.