Home > ESX, HP, VMware > Your HP blades may be underperforming… Why? From http://myvirtualcloud.net

Your HP blades may be underperforming… Why? From http://myvirtualcloud.net

Andre Leibovici, a fellow vExpert who works for EMC as a Sr. vSpecialist has blogged today about an issue with HP Blades and a power regulation setting where the default on HP Blades can cause performance issues.
http://myvirtualcloud.net/?p=1212

This is one of those great finds as it can be so difficult to troubleshoot “My VM is running slow” and you can go through your virtualisation environment with a fine tooth comb and find nothing as the issue is actually related to hardware settings.

Interestingly the issue manifests itself when the system is not fully utilised and CPU use is not high as Power Regulation only kicks in when it sees there are idle CPU cycles which makes it even more confusing to troubleshoot as you would be getting high CPU READY figures when CPU wasn’t stretched.

HP has a document, Power Regulator for ProLiant servers which explains it all.
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00300430/c00300430.pdf

The fix is to either disable Power Regulator on the hardware BIOS by setting it to Static High Performance mode which will disable any power savings or set it to a custom setting, OS Control mode and then configure ESX(i)’s Power Management Policy using the vSphere Client Host Configuration panel to High Performance to disable power management.

Thanks, Andre!

Categories: ESX, HP, VMware Tags: , , ,
  1. September 22nd, 2011 at 17:55 | #1

    It’s also possible to script these settings using HP’s RIBCL. I wrote a small article a while back about using PowerShell for looping thru enclosures and adjusting BIOS power settings via XML.

    http://damiankarlson.com/2010/10/26/modifying-hp-c-class-blades-via-ilo-and-powershell/

  2. WoodITWork
    September 22nd, 2011 at 22:35 | #2

    @Damian Karlson
    Thanks, Damian, nice addition.
    I really wish HP would write a native PowerShell module for Blades, OA and Virtual Connect so we don’t have to resort to Putty/Plink though.

  3. Pål Svendsrud
    September 24th, 2011 at 10:31 | #3

    We are using Unattended inst of SQL2008R2, to measure our VMware env. The job used to take 999 seconds – after the job just took 468 seconds!

  4. Rotem Agmon
    September 27th, 2011 at 06:30 | #4

    You may also follow the instructions provided in the following VMware KB:

    Poor virtual machine application performance may be caused by processor power management settings
    http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1018206

  1. No trackbacks yet.
 

Time limit is exhausted. Please reload the CAPTCHA.