CSE 60772 - High Performance Distributed Systems
This page contains account and cluster information that applies to all assignments. Read the individual assignment page for the work you must complete and the files you must submit.
Before doing anything else, follow the setup instructions to create your various accounts.
Anything to be turned in should be checked into your repository for this course,
which will grow and evolve as the semester develops. Please put each assignment
in a separate directory named according to the unit: warmup, simd, openmd, cuda and so forth.
In that directory, you should include all the technical materials that describe the experiments you ran, along with the results. Include these materials when they apply:
README.md file that addresses the questions and other objectives of the assignment.We strongly recommend that you add materials to the repository incrementally as you work, rather than trying to assemble things at the last minute. Commit as you go.
In all things, take time to organize your materials, give them constructive names, and make everything clear and consistent.
Once you have committed and pushed any necessary changes to your repository, simply go to Canvas and submit the assignment online by pasting the URL of your repository. Once an assignment is submitted, you are welcome to work on the next assignments by making commits in other directories.
If you are off campus, connect to the Notre Dame VPN before using SSH. Follow the Notre Dame VPN installation and connection instructions. If you are on campus, connect through the campus network.
CRC has two equivalent general front ends, crcfe01.crc.nd.edu and crcfe02.crc.nd.edu. You may use either one and switch to the other if your first choice is unavailable or busy. Replace NETID with your Notre Dame NetID and run one of these commands:
ssh NETID@crcfe01.crc.nd.edu
ssh NETID@crcfe02.crc.nd.edu
CRC also has condorfe.crc.nd.edu, a front end dedicated to HTCondor submissions. Use condorfe whenever you submit an HTCondor job.
All CRC front ends are shared by many users. A long-running or multicore program on a front end consumes CPU and memory needed by everyone else and can make logins, editors, compilers, and scheduler commands slow. Use front ends only for short, lightweight work. Submit long-running, CPU-intensive, memory-intensive, or multicore programs through Grid Engine or HTCondor.
After completing Set up SSH access to Anvil, run these commands after logging in:
whoami
hostname
mybalance
showpartitions
Your Anvil username will normally begin with x- and is different from your ACCESS ID. mybalance lists the Slurm accounts available to you.
The Anvil shell runs on a shared login node. Use it for short tasks such as editing files, compiling small programs, and submitting or checking jobs. Run long, CPU-intensive, memory-intensive, or multicore programs through Slurm so that they execute on compute nodes with reserved resources.