Quantum Control Cluster at the University of Oregon

Usage Policy:

When using the clusters, you must observe the following:

1. To zeroth order, there is no management staff for this cluster. Users must be self-sufficient, and should acquire competence with unix and high-performance computation elsewhere before requesting access to the cluster. There are various avenues for this, including physics/chemistry courses involving numerical computation, but ultimately research advisors are responsible for ensuring their students' readiness to use the cluster.

2. Be aware and considerate of other users, especially regarding cpu and disk usage. Translation: if there are many users on the cluster, don't go hogging the batch queue.

3. Ensure that you have a valid .forward file in your home directory. (This is created automatically when your account is set up, but you are responsible for keeping it current.)

4. All storage on the cluster is effectively volatile. There is currently no backup system in place for the cluster, so critical files should not be stored on the cluster without an off-site backup. Home directories are protected from hardware failures by a RAID system, but management will not make any attempt to recover lost data for any reason.

5. Likewise, compute nodes may occasionally need to be rebooted on very short notice; long jobs (say, 8 hours or more) should periodically checkpoint themselves to avoid loss of time/data.

6. The front-end nodes are for development and compilation only; long jobs should be run on the compute nodes. See the hardware page for details on the capabilities of the different compute nodes.

7. Use the batch system to run all computing jobs. You need to become thoroughly familiar with the batch system before using the cluster.

Accessing the Cluster:

• Log in via the main front-end node (qc.uoregon.edu, aliased to qcp on the private network). This is where you should edit files, compile codes, etc. All other nodes are accessible from here via the private networks. However, you don't really need to know much about the private networks, since you should be using the batch system to launch all your compute-node jobs.

• You may also log in via the storage node (qcs.uoregon.edu, aliased to qcsp on the private network). This is the preferred option for transferring large files to/from the cluster, so that it doesn't have to be relayed via the private network (assuming the files are in your home directory).

• Your home directory is located on the main RAID system, and is common to all nodes via NFS.

• If you change your password, do so only on qc.uoregon.edu; your password will eventually migrate to all the other nodes. Any logins via the private network (using ssh) do not require a password.

• You may use local disk space (/tmp, /scratch, etc.) on any compute node for fast, temporary data storage or checkpoints, to avoid network overhead in writing to /home.

• Report problems to Jonathan Mackrory (mackrory@uoregon.edu) or Daniel Steck (dsteck@uoregon.edu).

Obtaining an Account on the Cluster:

• In the initial phase of the cluster operation, access to the cluster is limited to participants of the NSF MRI grant that funded this cluster.

• Read the policy above, and make sure you have basic competence in using unix, the batch system, some programming language, etc.

• Print out this page.

• Print your name here:
  X__________________________________________

• Sign the following statement: I have read and agree to the above policy.
  X__________________________________________

• Have your research advisor sign the following statement: I acknowledge that this is my research student/postdoc/associate; I attest that this person is sufficiently competent with computation to satisfy the above policy; and I accept responsibilty for this person's actions on the cluster.
  X__________________________________________

• Bring this form to Jeremy Thorn (jthorn@uoregon.edu), 256 WIL.