Oct
15

Nutanix: When Blunt Force Trauma Isn’t Enough on the vSwitch

My philosophy with most things is to keep it simple. The more bells and whistles turned and tooted usually lead to problems down the road. Each Nutanix node has over 20 Gb/s of available traffic which is a truck load in my mind. Living in the VDI land it’s common space to lots and lots of desktops running on one host. Multiple desktops, multiple vMotion streams during maintenance mode can take most of a 10 Gb link. If you want to ensure performance of your desktops you should look at using a virtual Distributed Switch(vDS). Lots of improvements have been made over the years with vDS and a great resource is the WhalNetwork.

Nutanix has a requirement to have the vmkernel port for management and IP of the virtualized storage controllers to be on the same network. You can get around this by having multiple management vmkernel ports but I always push for the same subnet if possible. This in-part due to our current implementation of AutoPath. AutoPath allows our virtualized storage controllers to go down and the VM’s running on the host to keep on ticking, this in turn helps with rolling upgrades and ability to self restart services if need be without impact.

The key to the above message is that it’s same network not port group. I recommend to create a new port group just for the virtualized storage controllers and also create a specific User-defined network resource pool to go along with it.

The below images are from NX-2000.

vDS-1
Remember shares are only active when contention exists and there is actually load. Other words only the shares that have active traffic will be used to calculate the share value.
vDS-2

vDS-3

If you want further technical insight into Nutanix, check out the Nutanix Bible.

Speak Your Mind

*