Each 1MHz of CPU is defined as 50 units.

N= Number of CPU cores,   f= CPU frequency (MHz)

Total  cpuunit of HW node = N x f x 50

For example in vzlinux25: (2 x 2.4 GHz CPU cores)

Now if we run the following commands in HW node, we have:

# cat /proc/cpuinfo -à cpu MHz= 2400.112;  N=2, Total cpuunits = 2 x 2400.112 x 50 =240,011

# vzcpucheck à , Total cpuunits= 245,537 (close enough to the above calculation)

Cpuunits is the guaranteed amount of cpu that is assigned to a VPS. SP if we give 8000 cpuunits in the above case the MHz equivalent of it would be (8000/245537) x 2400.112 = 78 MHz

Currently this value is between 500 and 1000. Customer doesn’t see this parameter (Except in PSBM power panel which shows the value which is 80000 now)

—————————————————————————————————————————————————————————

cpulimit is the burst amount of cpu that is allocated to VPS/VM if resource is available and is a hard limit.

This is a bit more complicated and is explained in this KB: http://kb.parallels.com/en/112588

First we assume that 1 cpu core is assigned to the VPS so the maximum cpulimit would be 100% of this cpu core.

Again for a sample VPS within vzlinux25 we have:

cpu MHz(VPS) = cpulimit (VPS) x cpu MHz (HW node)

cpulimit = 60 %;

cpu MHz(VPS) = 0.6 x 2400.112 = 1440

By checking the cpu MHz in VPS we have:

# cat /proc/cpuinfo  (within the VPS)  à cpu MHz = 1439 (very close to the above calculation)

Now if we assign 2 cpu cores to the the VM/VPS the total would be 200% and the cpulimit could be any value between 0 and 200%.

This is the value that customer sees as the MHz of their cpu using the above command in the terminal or in the power panel.

So the oversell is already done by the difference between guaranteed and burst cpu.

The following table is the suggested values for plans and doesn’t match necessarily with the existing configuration.

The parameter  burst_cpu-avg_usage is always 100%