• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Dynamic core counts

L

lightp2

Gawd
Joined
Oct 28, 2009
Messages
672
I have been advised to use vCPU=1. I assume VMware ESXi scheduler logic is written to tune for vCPU=1, I do not know the detail but it is fine with me.

Is it possible ?

A. Configure and start VM vCPU=2 to obtain some benefits in special cases that I can manually verify results.

B. During low load time, can I manually disable VM cpu core thus reducing from 2 to 1. (I read one or two linux postings on disabling runtime physical cpu core, not sure whether this can be done when ESX is sitting in between). Will ESX benefits from this by reducing its runtime profile for this VM from vCPU=2 back to vCPU=1, thus meeting its generic scheduler preference setting for vCPU=1?

C. When the need arises, will reactivate the VM cpu core counts back to 2 temporary for some jobs, then deactivate back to 1 again after the jobs are completed, assuming all these are possible and ESX can accomodate dynamically ? yes/No?

Additional note: just to clarify. intend for all these adjustments made inside the VM while the VM is actively running

Before spending time/effort to test and find out results, would like to hear maybe others already have answers.
 
Last edited:
Update to my own question.

This test is not scientific at all, just based on visual inspection of several performance counters.

I did not install VMware tools for guest VM,

1. test done on VMware Server 2.0.2 on Windows 7 RC1. Yet to perform test on ESXi.

2. VM configured with vCPU=2

3. Live VM has cpu0 and cpu1. Deactivated cpu1, inside VM monitoring tool can see one CPU usage dropped to 0%

4. From Host Performance Monitor, the assigned 2-cores still had loads on them. very shortly after step 3, one core has higher usage while the other core maintained some usage percentile. When heavy usage inside VM, from host perfmon can see both cores usage % climbing, with one core higher than the other supposedly dormant core.

So initially from this observation, seems at least on the VMware Server side, it appears the dynamic cpu core deactivation inside VM will not trigger automatic adjustment in the VMware kernel runtime profile for this VM. (which means it still expects 2 vCPU)

OK looks like this idea of minor runtime core tuning will not work.
 
I see. all right I will test on ESXi next week then. Appreciate your kind clarification.
 
Updated Results

Host : ESXi 4.0 U1

VM-1 : F12i686
vCPU=2

again this is un-scientific result using visual inspection.

Determine effect of CPU adjustment

1. Inside VM1, runs hardinfo for simple cpu benchmark
2. Inside VM1, in the middle of benchmark, deactivate CPU1, system switched to UP mode.
3. In VMware VI client, look at the performance chart, clicking refresh a few times.
4. When the benchmark finished, manually inspect performance chart for the VM
4a. The basic screen still shows 2 CPU so from charting perspective it remembers this is 2 CPU VM
4b. more detail check against the CPU perf chart shows at some points CPU0 usage raised and CPU1 dropped to 1-MHz

I am not sure what kind of conclusion I get :), but it seems

1. System still remember as 2vCPU profile.
2. The scheduler seems able to dynamically reduce one-core to bare minimal (1-MHz).
3. So I suppose if needed, user can de-activate or re-activate CPU1 at will to meet computing needs.

Whether to do such arrangement and how much improvement in actual real-world condition depends on many factors.
 
You must log in or register to reply here.
Back
Top