Cluster utilization and self-provisioning

[TheVMinator]

I am moving to a cloud infrastructure with VCAC for self-provisioning. How does this impact my target utilization for my HA cluster?  Previously I was targeting running each cluster at 80% utilization of RAM and CPU on each host for average peak utilization.  now I am going to allow vms to be self-provisioned.  I won't control the provisioning process anymore but various clients and tenants can provision VMs at will without my notice.  As a result, I have to be able to have capacity available more quickly to add VMs, and not suddenly run out of cluster capacity.  I want to minimize waste by running my clusters to capacity, but I also need to maximize elasticity.  What are some guidelines on how to do this?  Anyone have experiences to share -

1. what did you pick as a target utilization figure and why ?

2. how did you capacity plan / forecast for cluster capacity?

3. did you use admission control?

[vbrowncoat · Accepted solution]

Sounds like a cool project. Keep in mind that from an infrastructure standpoint HA and admission control are still trying to solve the same problem, recover VMs from a host or OS failure as quickly as possible.

As an example, if your new cluster has 20 hosts and you want to be able to have a host in maintenance mode and still suffer a host failure and you've decided to use % based admission control policy (this is the default recommendation, I would recommend you evaluate your environment and determine if it is the right option for you), you'll want to set the % at 10%. This will ensure that your cluster has sufficient resources to restart all running VMs. Keep in mind that unless VMs have reservations, HA just reserves capacity to start the VM, there is no guarantee of performance.

As far as your target utilization, that depends on the SLAs you are providing and your tolerance for risk.

At the last customer I worked for the answers were:

1. We reserved capacity in a cluster such that we could have a host in maintenance mode and still lose a host and have no VMs experience performance degradation

2. vCOps

3. Yes

[Nick_Andreev]

@gs_khalsa,

Keep in mind that unless VMs have reservations, HA just reserves capacity to start the VM.

Does it? I always thought that percentage based admission control reserves resources only for VMs with reservations. And if there are no reservations, it doesn't reserve anything. Can you confirm or deny it?

[vbrowncoat]

See this post: Percentage Based Admission Control gives lower VM restart guarantee?‌

If no reservations are set:

Percentage Based will do the following:

1. The Percentage Based policy will take the total amount of resources and subtract the amount of resources reserved for fail-over. If that percentage is for instance 20% than 40GB and 4GHz are subtracted. Which means 160GB and 16GHz are left.
2. The reserved resources for every virtual machine that is powered on is subtracted from what the outcome of 1. was. If no reservation is set memory then memory overhead is subtracted, if the memory overhead is 200MB then 200MB is subtracted from the 160GB that was left resulting in 159,8GB being available. For CPU the default of 32MHz will be used.
3. You can power-on virtual machines until the amount of available resources, according to HA Admission Control, is depleted, yes many VMs in this case.