2024-12-22

2 thoughts on “VMUG UserCon Sydney: 2015

  1. Hi Keiran,
    Great article. Just a couple of minor corrections. vNUMA kicks in above 8 vCPU, i.e. when you configure 9 vCPU’s or more it will kick in. But in vSphere 5.5 and below vNUMA will not be active if Hot Add is also enabled. This will change in vSphere 6. You can change the default settings for vNUMA if you have say a 10 core per socket system, so that it provides the correct NUMA topology to any large VM’s.
    With regard to the number of VM’s per datastore in an HA cluster it is limited to 2048. So in Nutanix you can configure multiple containers, which are just NFS mounted datastores, and have as many VM’s in a cluster as the limits of the hypervisor allow. With vSphere 6 this will be 8000, with vSphere 5.5 it was 4000. So in the case of a deployment of 8000 VDI Desktops in vSphere 5.5 you could have had 2 vSphere clusters of 32 hosts each on a single 64 node Nutanix cluster (in just 32 RU) and used only 4 datastores in total for all of the VM’s. This greatly reduces management overheads. Note there is no technical limit to how many nodes can be deployed in a single Nutanix cluster but there are practical considerations around failure domains and hypervisor limits. But you are not limited to one hypervisor cluster per Nutanix cluster.
    Glad you enjoyed the presentations. The slide decks will be posted up to the VMUG site for everyone to get a copy of.

    1. Hi Michael.
      Thank you for ready and thank you for the corrections. I have updated in regards to vNUMA. Great information in regards to Nutanix! I’ll update it shortly.
      I am keeping an eye out for the slide decks, working from my short list of notes, was a great day to remember.
      Thanks.

Leave a Reply

Your email address will not be published. Required fields are marked *