April 9, 2014 - Sven Huisman

Don’t forget the BIOS power management setting of your VDI-hosts!

I sometimes come across a VDI environment where a certain setting is not set according best practices. The setting I refer to is a well known best practice, but is still often forgotten. The setting can be set two different ways and I discovered that one way will not (always) work. I’m talking about the Power management setting of the ESXi host. The best practice for VDI-hosts is to set this to “High Performance”.

Recently I was at customer and they use an application to do certain calculations. They had a sample script to do the a couple of calculations. On a VDI-desktop on one host this task took 2 minutes to complete and on another host this took 1 minute. This behaviour was consistent. I found out there was a difference in Power Management settings. On the host where the calculations took 2 minutes, the Power management in the BIOS was set at default “HP Dynamic Power Savings Mode”:

HP2

On the host where the calculations where completed in 1 minute, the BIOS was set at “HP Static High Performance Mode”, no reboot required:
HP3

This is the prefered setting! Once I set this on the other hosts, the calculations took 1 minute on each host.

There is one other way to set this on “High Performance”. If the power management setting in the BIOS is set to ” OS Control Mode”:

HP1

You can set the Power management setting with the vSphere client on the host:HP4

 

 

However, I found out that this setting does not (always) work, or at least in ESXi 5.5 on HP server hosts it doesn’t. So the preferred method is to set this to Static High Performance in the BIOS. I’m sure Cisco, Dell, IBM and the other vendors have similar settings.

Virtual Desktop

Comments

  • PiroNet says:

    This is a very valid point you made there.
    Some server vendors can also let you configure PCIe power management in their BIOS.

    If it happens it is set to anything else than high performance you might end up with strange issues with your IO cards.

    For instance we had this server with PCIe power management set on enable which had serious issues with Intel 10GbE cards that would not ‘wake up’ and totally frozen (but would not failed and port remains up) when PCIe slot power changed.