[pve-devel] intel pstate: wrong cpu frequency with performance governor

Alexandre DERUMIER aderumier at odiso.com
Tue Sep 20 11:20:20 CEST 2016


>>What is the suggestion - disbale the pstate driver? I think that would increase 
>>power consumption? 

>>yes, this is increase power consumption. But increase stability & latencies.

Note that disable pstate, don't disable dynamic frequencies.

we still have classic "cpufreq" governors.

with pstate, we have 2 profiles:
-powersave
-performance 

cpufreq have

-powersave
-ondemand
-performance 


What I don't understand
"
https://www.kernel.org/doc/Documentation/cpu-freq/intel-pstate.txt

"Since these policies are implemented in the driver, they are not same as the
cpufreq scaling governors implementation, even if they have the same name in
the cpufreq sysfs (scaling_governors). For example the "performance" policy is
similar to cpufreq’s "performance" governor, but "powersave" is completely
different than the cpufreq "powersave" governor. The strategy here is similar
to cpufreq "ondemand", where the requested P-State is related to the system load."
"


for me, pstate "performance is working like cpufreq "ondemand".



----- Mail original -----
De: "aderumier" <aderumier at odiso.com>
À: "dietmar" <dietmar at proxmox.com>
Cc: "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Mardi 20 Septembre 2016 11:05:25
Objet: Re: [pve-devel] intel pstate: wrong cpu frequency with performance governor

>>What is the suggestion - disbale the pstate driver? I think that would increase 
>>power consumption? 

yes, this is increase power consumption. But increase stability & latencies. 

my old xeons (previous sandybridge) and amd, always are at maximum cpu frequency. 

Redhat have a special daemon "tuned", which have some tunables profiles for this (latency,bandwith, ...) 


>>Really? You observed retransmits because of different CPU frequency. I think 
>>retransmits need 
>>much larger speed differences, for example caused by high load (as opposed to 
>>idle nodes). 

I'm seeing a lot less retransmit since I have disable pstate. 
(but maybe this is because of bug where frequency was stuck low, and I have also a big cluster with a lot of vms) 


----- Mail original ----- 
De: "dietmar" <dietmar at proxmox.com> 
À: "aderumier" <aderumier at odiso.com>, "pve-devel" <pve-devel at pve.proxmox.com> 
Envoyé: Mardi 20 Septembre 2016 09:52:51 
Objet: Re: [pve-devel] intel pstate: wrong cpu frequency with performance governor 

> I think this is the normal behavior of pstate, as they are a lower limit. 
> 
> But for virtualisation, I think it's really bad to have changing frequency. 
> (clock problem for example). 

What is the suggestion - disbale the pstate driver? I think that would increase 
power consumption? 

> Also for corosync, that mean that a loaded node will have faster frequency, 
> and non-loaded node low frequency. 
> This can give us reload corosync.conf after removing a node., because low 
> frequency take more time to get the corosync message than the fastest node 
> https://www.hastexo.com/resources/hints-and-kinks/whats-totem-retransmit-list-all-about-corosync/ 

Really? You observed retransmits because of different CPU frequency. I think 
retransmits need 
much larger speed differences, for example caused by high load (as opposed to 
idle nodes). 

_______________________________________________ 
pve-devel mailing list 
pve-devel at pve.proxmox.com 
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 




More information about the pve-devel mailing list