Listed below are instructions for setting up accounts on MARCC/BlueCrab located at JHU (with co-support from UMD).
Instructions:
If asked to form an account on MARCC you need to request to be added as a user. Use the following link:
http://www.marcc.jhu.edu/request-access/request-an-account/
Once this is done, Dr. Klauda will need to approve this request and with a day you should have an account.
Details of MARCC can be found:
http://www.marcc.jhu.edu/about-marcc/about-us/
Now for the cluster setup:
If you are not familiar with the use of LINUX, I would first go through some on-line tutorials for commands to use in LINUX and here (Linux Guide). To login to MARCC you will need multi-factor authentication:
http://www.marcc.jhu.edu/getting-started/connecting-to-marcc/
Windows you will need to have PuTTY (Extra Software) to connect to MARCC:
http://www.chiark.greenend.org.uk/%7Esgtatham/putty/
And the following program will be useful to transfer files between MARCC and your computer
http://winscp.net/eng/index.php
Built-in commands for MAC users can be used to connect via the command line (scp) and apps for file transfer (eg, cyberduck).
Then, take a general look at the website of the MARCC to get a general idea on how to login and submit jobs via SLURM. Your jobs should not run in your home directory. They should instead be in the work directory. This is a linked directory in the home. It is shared with all users so please make a user directory with your user name (replace username with your username (w/o @umd.edu):
cd work mkdir username
Example files for running on gpus
Make adjustments depending on your system
marcc-prod.gz
Example file for running on CPUs
Make adjustments depending on your system
dyn.zip