Thank you for all the feedback on my previous two posts on how vCenter CapacityIQ can identify Idle and Powered-Off VMs and Oversized and Undersized VMs. I wanted to follow up with a screenshot I got from a recent customer POC:
Almost 95% of the virtual machines were oversized!!! I believe the VI administrators out there are perhaps not as surprised as I was. Looks like every application owner wants a large VM, ending up with lots of over-provisioned VMs. Hopefully CapacityIQ can now provide VI administrators a concrete evidence that can help them justify why such oversized VMs should be rightsized.
6 comments have been added so far
Out of curiosity, for the oversized and undersized VM report, was it using the default out of the box CapIQ settings? By default the oversized and undersized is set at just 1%. I’m wondering was this tweaked for another value that was used in this customer’s POC?
I would be interested in the answer to this as well. If the defaults were changed – what is the best practice to set theses defaults to?
I believe the screen shot here is not details enough. Oversize or not is really depend what is the usage of the current infrastructure and the growth of the system been provision in the environment. It is very subject to the environment for an IT administrator who manage it.
@William, @Maish – This was based on default settings.
@Craig – CapacityIQ uses configurable thresholds to compare to the actual infrastructure usage as well as capacity buffers that you can tweak. Agree that it’s environment specific.
Can someone please specify what is Best practice to use for NON-DEFAULT thresholds of Undersized and Oversized VMs???
“Best Practices” wont really help because every IT shop is different. For instance, if you do local agent-based backups, you’ll see a CPU/Mem spike during that time and you’ll have to account for that in your Global Settings.
If you just looking for an example, here’s ours.
Oversized % = 75%
Oversized CPU/Mem% = 30%
In other words, if the CPU or Mem is utilized less than 30% over 75% of the timeframe, the VM is too big.