This page is a primer on the Condor High-Throughput Computing System. Below, you will find information about what Condor is, how to connect, and what kinds of work you can do on the system.
What is Condor?
The Condor system is an implementation of the HTCondor job scheduling system available to all RCC users as a free resource.
What is it used for?
Condor is used to run larger batches of jobs that are typically smaller or shorter than jobs submitted to the HPC. These jobs may take minutes or may be as long as days to weeks. When deciding whether to use HPC or Condor (HTC):
- Use the HPC if you need to run a large, long running parallel batch job as efficiently as possible in a given time-span
- Use Condor if you need to run a large number of serial batch jobs as efficiently as possible in a given time-span.
Benefits of Condor vs HPC
- You can run hundreds to thousands of serial jobs at once. * You typically do not need to recompile your code to take advantage of Condor resources.
What Condor cannot do
- Parallel applications using OpenMP or MPI libraries will not work on our implementation of Condor. Use HPC for these types of applications.
- Condor is a shared-nothing system, meaning each job has access only to its own memory. Jobs cannot make remote system calls.
For more general information about HPC vs HTC, refer to the official documentation.
Connecting to Condor
You will also need to connect to the FSU VPN if you are off-campus.
To submit jobs to Condor, connect via SSH to the HPC login nodes:
The Condor submit node includes software that allows you to submit jobs to the Condor queue, compile software for Condor, and use common software available on the system.
Things to Keep in Mind When Using Condor
There are some things to keep in mind about the FSU Implementation of HTCondor:
- Condor is free for all RCC users. There is no concept of ownership in the Condor system, and all users are given the same level of access to compute resources.
- Condor utilizes a fair share queuing system. The system will automatically prioritize jobs according to the amount of resources an individual has already used. The more resources an account uses, the lower the priority for that user's future jobs. This ensures all users get access to Condor resources over a period of time.
- Condor consists of hardware that has been donated or recycled from other RCC systems.