I've been experimenting with Amazon EC2 and SharePoint to better understand how an enterprise architecture would be built on top of it as a platform and, as a part of this exercise, better understand the cost structure.
One key number that escaped me during my initial review of the pricing structure is the one highlighted below:
Now before we get into the details of what to watch for here, I should preface and say that for use cases that demand high performance disk I/O, provisioned IOPS is probably well worth it and, in fact, I definitely foresee us using this in select scenarios (i.e. high I/O database server instance).
In the screen cap below, you can see that I've got three provisioned IOPS volumes configured with 2000, 2000, and 1000 IOPS respectively (note that maximum IOPS on a volume is a multiple of the size of the volume):
The question is what does "$0.10 per provisioned IOPS-month" actually mean?
Well, here's the skinny from the Amazon web site:
For example, if you provision a volume with 1000 IOPS, and keep this volume for 15 days in a 30 day month, then in the Virginia Region, you would be charged $50 for the IOPS that you provision ($0.10 per provisioned IOPS-Month * 1000 IOPS Provisioned * 15 days/30).
So in my case, for a 30 day month, it would work out to be (2000 + 2000 + 1000) * 0.10 * 1 = $500/mo. for the provisioned IOPS volumes.
This is something to keep your eye on as if you are using EC2 instances with provisioned IOPS volumes sporadically, you may want to take snapshots or create an AMI and discard the volumes when not in use. Comparatively speaking, snapshot data is much friendlier on the wallet when your system isn't under active use.