So after a week of VMworld 2011 goodness, I think everyone has heard at least a little bit about VMware vCloud Director resources allocation models:
Pay-as-you-go: Resources are committed only when new vApps/VMs are created. Unpredictable resource consumption/utilization/cost. EX: Amazon EC2
Allocation Pool: Only a percentage of the resources are committed. Only a percentage of those resources are guaranteed. Fairly predictable resource consumption/utilization/cost. However, allows for bursting.
Reservation Pool: All of the resources granted to an OrgvDC are committed at the time that the OrgvDC is created. 100% of the cluster resources are guaranteed.
We won’t go into the pros/cons of each of these allocation models in this post. However, I did want to show you some screenshots of what happens when you consume all of the cluster resources — when using Reservation Pools.
Let’s take a look at a standard Reservation Pool model OrgvDC — at the time of creation:
Nothing unusual here — <this is just a small 2-ESX node cluster>
However, now let’s consume the remaining Cluster Resources by creating a new OrgvDC with all the dials set to ’11′
Note: I am going to consume ALL of the remaining vSphere (or cloud) resources. After clicking the [OK] button, all went well — the new “Monster OrgvDC” was created successfully.
Now, Let’s see what happens when I try and create just “1″ more OrgvDC: <Drum Roll>
As you see here, with NO remaining cluster/cloud resources available, vCD WILL NOT allow you to create new OrgvDCs. Kinda makes sense though. Allocation Pool and PAYG models are actually disabled.
If I had actually done the “math” right, I think Reservation Pool models would have been disabled as well. In this particular test, I did have a VERY small amount of resources available, which vCD will let me commit. However, in real life, these settings would pretty much make VMs unusable.
Just as an FYI — Here’s what the Provider vDC [Monitor] tab results show:
This just gives you a look at what vCD does when you consume ALL of the resources when using Reservation Pool models. If you are keeping an eye on things, this should not be a problem — you’ll just need to add additional servers as your demand for capacity grows.
“Be careful with that Reservation Pool, Eugene…”
@wsellers
