MUSes Computational Policy


Users may run long computation jobs on up to 3 computers at a time. If the user needs more than 3 computers, the user must get permission from the system administrator. The web page www.math.uiuc.edu/help/policies/longjobs will show users who have requested such permission and the duration of their jobs.

When computationally intensive classes are using the lab, the total number of lab computers minus six, divided by the number of students in all the classes will be the number of computers available to each student for computation. Students may negotiate among each other to get unused computers from their classmates, but if they do so, they must report it to help so that it can be noted in www.math.uiuc.edu/help/policies/longjobs.

On the compute servers (u23 and u24) there should never be more computationally intensive jobs running than there are processors. u23 has 2 processors and u24 has 4 processors. This means that there should not be more than 2 computationally intensive jobs running on u23 and there should not be more than 4 computationally intensive jobs running on u24. If a machine has already reached its maximum number of computationally intensive jobs, DO NOT begin additional such jobs on that machine.

All jobs run on lab machines (currently u01-u40) should be run at a low priority. This means that if a user is working at the terminal, they will be able to get work done, but that if no one is logged in at the terminal, then your job will get as large of a fraction of the cpu as if it were at normal priority. For help with setting the priority of a job, consult the Frequently Asked Questions.

No computations are to be run on ssh.math.uiuc.edu, aka u00.math.uiuc.edu. Please run jobs requiring less that 50 mb ram on u15 - u22, jobs requiring less than 150 mb ram on u01 - u14, jobs requiring less than 400 mb on u25 - u40, and jobs requiring more than 400 mb of ram on u23 (2x296 MHz UltraSPARC-II processors and 2 gb ram) and u24 (4x296 MHz UltraSPARC-II processors and 2 gb ram). Please nice all jobs and don't run more jobs than processors. Use top to check before running.