NCSA Home
Contact Us | Intranet | Search

ncsa

Previous: User_Assistance
Next: GNU_Utilities
View by Date; View by Name

Interactive Limits

May 15, 1998
 
As of May 18th 1998, an automated interactive usage policy will be in 
place. User limits (for all active login sessions) will be as follows:
- a total of 8 processes 
- a total of 512Mb memory
- CPU time of 15 mins per process/1 hour per job 
 
A "job" is defined as what gets run when a user executes a command (for
example, a program executable). This is implemented as a "process group".
 
Violation messages are sent to the user's active terminal windows along
with the list of processes killed. The message will be as follows:
 
Job CPU time violation:
  ** VIOLATION: USER(lahti) Process Group Time(1:13:31) exceeds allowed 
  limit(1:00:00) **
 
Process CPU time violation:
  ** VIOLATION: USER(lahti) Process Time(15:23.92) exceeds allowed limit
  (15:00.00) **
 
Total number of processes violation: 
  ** VIOLATION: USER(lahti) CPU Load(7.87) exceeds allowed limit(7.50) **
 
Memory violation:
  ** VIOLATION: USER(lahti) RSS Load(720.23M) exceeds allowed limit(512.00M) **

Note that the limit monitoring system enforces process and memory limits based 
on total "cpuload" and "rssload". These are defined as follows:
 
  The cpuload for a process is computed as the ratio of cputime to wall 
  clock time over an interval. Currently, a 30 second interval is used.  
  The total cpuload for a user is the sum of the cpuload for every 
  interactive process that the user has running (during the interval).
 
  The rssload for a process is RSS (resident set size) * cpuload. 
  The total rssload for a user is the sum of the rssload for every 
  interactive process for the user.

 
Please use the batch system for larger jobs. See /usr/news/lsf for details.

Silicon Graphics Origin2000:usr/news/Interactive_Limits
Last Modified: May 15, 1998