Exploring the difference between effective capacity and utilization

Alex McDonald writes The Missing Shade of Blue blog for Netapp. A week or so ago, Alex started a blog dialogue about disk utilization and capacity metrics where he was critical of 3PAR's disk utilization claims in our most recent SPC-1 benchmark.   I disagreed with Alex in a video and then we were off to the races with the usual blog smackdown chain of events.

Then a funny thing happened.  We found a common ground by identifying areas where the terms we were using weren't working so well and we decided to work on that to clarify things.  It seemed like a bit of a breakthrough and was refreshing in that blogging detente happened.

In a nutshell, we agreed that:

1) "Disk utilization" should represent the device level view of the world.  How full are the raw devices that are being used?  For instance, if you mirror disks, they can both be 80% full and the disk utilization would also be 80%.

2) "Effective Capacity" should represent the application view of the world.  How much file system or database space is there to work with as a percentage of raw disk capacity?  Using the example above with disk mirroring, the effective capacity would be 40%, because a file system only "sees"  50% of the raw capacity in the disk mirror.

Here's a breif video discussion, but the blog comments do a much better job of detailing the arguments and the resolution.

Comments

[Alex McDonald]

Thanks for blogging on this, Marc.

I'm not 100% agreed -- yet. And I've run out of time for more this week; there's a lot going on at the moment, so it's on the back burner for a while.

Back atcha.

[marc farley]

No problem. We made good progess.

[Pete Steege]

I like your definitions of Disk Utilization and Available Capacity. I would go further and call the latter Available Content, since it's the amount of unique information, not the storage, that matters at that end.

Love the floating camera! How did you do it?

[marc farley]

That's a good point Pete. I agree. Calling it content instead of capacity clarifies it further.