Please provide your feedback in this short Flings' survey.
Oct 12, 2020

HI Mark,

Appreciate everything that you are doing here.. Trying to run VMware Compute Optimizer (VMCO). But couldn't be able to find the Get-"OptimalvCPU_v2.1.0.ps1" file anywhere.. Could you please help us here..

Oct 19, 2020

Hi Ck, sorry for the slow reply, as I was on vacation last week! Make sure you have both .ps1 files in the same directory when you launch the "Virtual_Machine_Compute_Optimzer_v2.1.0.ps1" script. It will look for it in the same directory.

Jun 24, 2020

Confused on the Medium assigned based on the "Host hardware in the cluster is inconsistent. Consider sizing VMs based on the minimums for the cluster".
The VM is 2core x 1socket . The recommendations are the same core/socket mix, so nothing to change, but still Medium?

Jun 25, 2020

Hi jbg. That was the recommendation that Mark A and I came up with for that scenario. Since we don't report on clusters individually, it's quite possible that no VMs would be misconfigured, and we wouldn't be able to flag the cluster inconsistency. We have it throwing that error for every VM in an inconsistent cluster at this point, but I've toyed with the idea of checking DRS (no DRS, no chance of a VM migrating to a host that might cause issues) and only comparing VMs in a cluster against the lowest host in the group.

Thanks for your feedback, and I will let you now if/when we update how that reports!

Jul 08, 2020

Hi Mark G. That change would be great to have. I have a main cluster that has 3 different Dell R9x0 class servers, so I am getting a lot of the same msgs as jbg mentioned.

Either an updated script that gets the lowest server config in the cluster or maybe a script option that allows us to specify the settings manually inside the script?

Jun 19, 2020

Our DBA says that with our current 1 socket/64 core setup he is seeing MS SQL correctly auto-configure itself for 4 NUMA nodes. Per this script, we should be using 4 sockets/16 cores. Is our performance being hurt on the VM side since it appears the application is working as expected? Is this some magic that is only applicable to MSSQL 2016?

Jun 22, 2020

To clarify, how much memory does the VM have, and how is the host configured (cores/sockets and memory)? I'll bounce this off of Mark A's performance team for the SQL side to see what their recommendations are in this scenario.

Jun 22, 2020

Thanks, Mark! The ESXi host is an HP DL580 with 1TB of RAM and 4 intel Xeon E7-4850v4 CPU's (128 logical CPU's in total). The SQL VM has 64 vCPUs and 768GB of RAM.

Jun 23, 2020
Jun 25, 2020

So, I bounced the scenario off of Mark A, and this was his reply:

So - yes - if the VM was built 1x64 under vSphere 6.5 or later it will show 4 vNUMA nodes since that is configure automatically now – BUT – the cpu topology is not correct. It should be 4x16 to be optimal for Windows scheduler.

I hope that helps. Please let me know if you have any more questions! Thanks for your feedback!

Jun 19, 2020

It looks like you don't check for the different NUMA options on the AMD platform.
I was looking for something that will be able to determine fast the status of the NPS setting and the CCX as NUMA option at each host in my AMD clusters.

Jun 22, 2020

Hi Zibi, I'm looking into whether or not the NPS/CCX settings are exposed to the host, and can be queried through PowerCLI. I'm hoping there is a way, but haven't find any documentation on the topic so far.

Jun 23, 2020

Hello Mark.
I keep my fingers crossed.
It would be shame if such fine solution could not provide the guidance for the real NUMA jungle in the AMD Rome environments.