Guidelines for Sharing CSLab Machines

CSLab has a limited amount of equipment, especially servers, which we must share fairly. Please keep the following guidelines in mind:

  • Application servers (e.g apps1.cs) exist to provide a functional environment to hundreds of remote users, and provide a server for standard applications such as email, web browsers and the like. If you have computing jobs that will consume considerable resources (memory or cpu) please run them on the compute servers (e.g. comps0.cs) as these machines are designed for that purpose.
  • Do not run more than one long running job at a time on any machine. If you have a number of them to run, put them in a shell script and run them one after another.
  • Use the nice command to lower the priority of long running jobs. If you use tcsh or csh, run your job like this:

nice +19 command args... &

If you have forgotten to use nice, you can use the renice command to lower a job's priority after it has started. Regardless of the effect the nice command has on the process, it also serves as a way to identify to system staff that the process is deliberate, and is not a runaway.

  • Monitor your jobs to make sure they don't run into infinite loops or start growing without bounds. Use the ps or top commands to keep an eye on them. You can also use the alarm command to limit the amount of elapsed time or CPU time that a process may take up.
  • Your home directory is in a pool of space that is shared by other users, usually people who are in the same group or have the same supervisor. There are no quotas limiting the amount of space an individual user may use. Please run the hogs command periodically to monitor your disk space usage, and remove old files when you no longer need them. If you run out of disk space, the sponsor of that disk space pool can add more space for a fee. Contact your Point of Contact (PoC) to discuss the options.
  • Do not put content on your personal web page that will generate enormous load on our web server (e.g. files in great demand on the web).

For the sake of other users, system staff may have to stop or kill jobs, or otherwise limit usage, when the guidelines are not being followed.