mjhjmtu

There are 32 cores in a system (2 cpus each 16 cores) and I want to
limit a user to use only 4 cores (100% usage for that 4 cores). There
are two kernel parameter for that according to the manuals:
cpu.cfs_quota_us and cpu.cfs_period_us

An example on the kernel.org document says:

 With 500ms period and 1000ms quota, the group
 can get 2 CPUs worth of runtime every 500ms.
 # echo 1000000 > cpu.cfs_quota_us /* quota = 1000ms */
 # echo 500000 > cpu.cfs_period_us /* period = 500ms */

I want to know how those ms are determined? In my case, I think the quota is 32000000 and the period is 4000000. Am I right? The content of etc/cgconfig.conf is:

group ansys 
 cpu 
 cpu.cfs_quota_us = 32000000;
 cpu.cfs_period_us = 4000000;
 
 memory 
 memory.limit_in_bytes = 8000m;
 

But it doesn't work! and I get an error when I run /etc.init.d.cgconfig start, I get this error:

Failed to parse /etc/cgconfig.conf or /etc/cgconfig.d
[FAILED]Starting cgconfig service: /sbin/cgconfigparser; error loading
/etc/cgconfig.conf: Failed to remove a non-empty group

Have a look at cpuset cgroup/container parameter. It lets you pin containers to specific cpu cores. In your case, you can specify only a single container to cores 0-3, and let no other container execute on the same core, there by allowing 4 cores with 100% utilization per container.

Furthermore cpuset is a more effective distribution mechanism when compared to cpu (that uses shares/quota).