The goal of our computing usage policy is to meet the needs of a diverse community as fairly as possible and to encourage parallel computing. We will periodically evaluate the policy and make adjustments as experience and feedback warrant.
The login, collaboratory, and compute nodes are described in detail in
CTC Computing Resources.
Fileservers
Windows: \\tc.cornell.edu\tc
Linux: /home/nfs/ctcfsrvn/m
Inappropriate or inactive files should not be stored on CTC file servers. This includes:
- Files that are not relevant to your CTC research project
- Unlicensed proprietary files
- Files associated with an account that has been inactive for longer than one year
If inappropriate or inactive files are found, we will contact you. If you then do not remove the files, we will remove them.
Note: CTC staff would not normally browse through users' files, but may be required to do so to resolve systems issues or consulting problems.
Login nodes
Windows: ctclogina, ctcloginb, ctcloginc
Linux: linuxlogin1, linuxlogin2
Everyone with a CTC account can log into any of the login nodes using one of the methods described in
Accessing CTC Machines.
The login nodes are for job development only. You can edit and compile programs, access files on the file server (H:), submit and monitor batch files, and view your job output.
Restrictions: Do not run jobs on the login nodes, because the machines are not designed for heavy cpu loads and response times for anyone using the node will become extremely sluggish. Debugging requires running the job, which should be done in the Collaboratory.
Collaboratory nodes
The collaboratory is located in 651 Rhodes Hall. You can develop and debug directly on a collaboratory workstation, as compared to the login nodes, which are accessed remotely. They are also appropriate for running jobs that are interactive or have a graphical user interface (GUI). More information about the collaboratory, including hours, is in Collaboratory Availability.
Restrictions: Do not lock a machine and leave it unattended. This deprives other users of accessing the machine. If someone else needs a machine that has been locked, the login session will be cancelled. If there is data that has not been saved, it will be lost.
Compute nodes
Accessible only via the Velocity Scheduler (vsched).
There are two main types of restrictions listed for the compute nodes.
Time limit specifies the maximum time you can request for any single job as specified in the .xml file with the <minutes> tag. (Note the .xml file specifies minutes, and the limits are in hours.)
<?xml version="1.0" ?>
<!-- Sample XML Job File -->
<job>
<nodes>1</nodes>
<minutes>300</minutes>
<type>batch</type>
<affiliation>vplustest</affiliation>
<run>\\tc.cornell.edu\tc\users\your_userid\MyJob.bat</run>
</job>
Waiting limit restricts the jobs a userid can have waiting in the queue for the specified node type. The limit is hours multiplied by nodes summed for all jobs submitted by that userid. It does not include any jobs that are running, only those that are waiting to run. (Note the .xml file specifies minutes, and the limits are in hours.)
hours is from the .xml file: <minutes>300</minutes>
nodes is from the .xml file: <nodes>24</nodes>
To view detailed resources limits, issue "vsched -policy" from a login node.
Development Cluster ("development")
- Waiting limit: 3 hours
- Time limit: 20 minutes
Serial Nodes ("serial")
- Waiting limit: 300 hours
- Time limit: 24 hours
Velocity 3 Cluster ("v3")
- Waiting limit: 3840 hours
- Must run at least 2 node jobs
- Time limit: 24 hours
Linux Development Cluster ("v2linuxdev")
Linux cluster: ("v2linux")
