[Engine-devel] Supposed to add per CPU usage related infomation into engine core and vdsm

Gilad Chaplik gchaplik at redhat.com
Mon Mar 10 11:36:20 UTC 2014


----- Original Message -----
> From: "Chuan Liao (Jason Liao, HPservers-Core-OE-PSC)" <chuan.liao at hp.com>
> To: "arch" <arch at ovirt.org>, "vdsm-devel" <vdsm-devel at lists.fedorahosted.org>, engine-devel at ovirt.org
> Cc: "Gilad Chaplik" <gchaplik at redhat.com>, dfediuck at redhat.com, msivak at redhat.com
> Sent: Monday, March 10, 2014 11:17:13 AM
> Subject: Supposed to add per CPU usage related infomation into engine core and vdsm
> 
> Hi All,
> 
> In order to support NUMA and guest NUMA feature in ovirt.
> We need per NUMA node CPU usage or per CPU usage related information on
> engine core.
> 
> This information will be used for VM who have NUMA aware information and find
> the best Host to run it on with best performance.
> 
> Approach 1:
> 
> 1.      Sample data in vdsm for each CPU stats. (sys, usr, iowait, idle)
> 
> 2.      Calculate the each CPU usage information. (%sys, %usr, %iowait,
> %idle)
> 
> 3.      Transport the usage data to engine core.
> 
> 4.      Engine core merge per NUMA node CPU usage. (%sys, %usr, %iowait,
> %idle)
> 

+1, since this data can be used for hardware other than NUMA

> Approach 2:
> 
> 1.      Sample data in vdsm for each CPU stats. (sys, usr, iowait, idle)
> 
> 2.      Calculate the each CPU usage information. (%sys, %usr, %iowait,
> %idle)
> 
> 3.      VDSM merge per NUMA node CPU usage. (%sys, %usr, %iowait, %idle)
> 
> 4.      Transport the usage data to engine core.
> 
> Which one do you prefer, and why, or other solution.
> 
> Best Regards,
> Jason Liao
> 
> 



More information about the Devel mailing list