HPC New User Documentation
This page is obsolete
Click the menu to the left for information on various topics:
- HPC News Blog to read about announcements and other developments.
- Status to check the current status of the cluster and the near-line storage.
- User Policies to review official UK Policies and Procedures.
- Account Information for information on an HPC cluster account.
- Hardware for a description of the cluster hardware.
- Documentation for other documentation.
- FAQs for a list of Frequently Asked Questions.
Computing Acceptable Use Policy
Users of any University of Kentucky Computing or Communications facility are expected to abide by all of the relevant policies and procedures. For more information, please see the Policy Governing Access To and Use of University Information Technology Resources (PDF) document or the UK Information Systems Policies page.
Accounts are granted to one person only. No other person may use your account. No exceptions will be made. Failure to abide by these policies may result in the suspension of your computing privileges at the University of Kentucky or other disciplinary action.
Other useful information
- Please forward email from the HPC cluster to an email address that you read regularly! This will allow us to inform you of any important announcements or problems with your jobs. Create a file in your home directory named .forward (note the leading period). This file should contain a single line, which is the address to which your mail should be forwarded.
- All interactive jobs must be run on a login node. No interactive jobs may be run on any other node.
- The HPC system is a cluster of nodes interconnected with a hi-speed network. In general, the primary approach to using more than 12 cores (on the Basic nodes) or 32 cores (on the Hi-Mem nodes) for one job is to use MPI. For more information, see the MPI documentation in the Documentation section of this Web site.
- Gaussian is not configured to provide multi-node execution. Therefore, the maximum number of cores for a Gaussian job is 32 . Gaussian must be run as a batch job. To learn about running Gaussian batch jobs using a helper script, read executing Gaussian as a batch job in the software documentation section of this Web site.
- Your home directory (/home/userid) is backed up nightly. Your scratch space (/scratch/userid) is NOT backed up, even though there may be a link to it in your home directory (/home/userid/scratch) for convenience.
The disk space used by each user's home directory is limited by quotas. There are no quotas on your scratch directory: (/userid/scratch).
When you reach the soft quota limit you will start getting warnings. You should immediately start reducing the amount of disk space you are using by deleting unnecessary files, and so on. You may move files off the cluster to the HSM near-line storage system. For more information on HSM and data storage, see the FAQ Long Term Storage.
Please use disk space frugally, both home and scratch! We don't have enough space for everyone to use the full quotas on their home directories, and everyone on the cluster shares the scratch space.
Please create any large temporary files in your scratch directory. This will give your jobs better I/O performance. The path for the scratch area is /scratch/your_userid.
Note: Files in the scratch directory are not backed up! Make sure you copy any important files from scratch to your home directory, the HSM near-line storage system, or some other safe location.
Files left unused in the scratch directory for more than thirty (30) days may be purged. This time limit may be adjusted due to extraordinary circumstances, such as when the file system nears full capacity. Every effort will be made to retain files for at least thirty days.