NIce to have you drop by Robin and you most likely took what I wrote out of context once again. In using raw flash, the intent wat that it was void of the components required to fit it into the SSD category. This reminds me of much the same when someone argues GiB over GB as I am certain you were well aware of my intent. The conversation we should be having, really, is where the resources are coming from to fuel enterprise storage arrays as they don’t have the componentry or software/firmware to complete the job. Although these storage arrays may be ideal for the big dogs, what about everyone in that middle ground that enterprise SSDs are only just starting to feed now?

Again…thanks for dropping by and glad we could help your blog out a bit!

No one is talking about using “raw flash” in an enterprise storage array. Everyone realizes that flash needs a controller, an FTL, EDC, garbage collection and the like. “Raw flash” isn’t on the menu.

The issue is this: is a 2.5″ SSD form factor that comes with the software overhead of SATA or SCSI and all the electrical components of a disk – components that are responsible for ?50% of all drive failures – along with other issues such as space and power efficiency, the right device for building enterprise storage arrays? Or is there a better way to use flash in enterprise storage arrays?

Several companies – including Oracle – are offering high-end flash arrays that use flash in form factors other than an SSD so I’m not the first person to question the value of SSD-based arrays. I hope this explains why I found your comment incomprehensible and not worth publishing.

I’ll be publishing a post soon that looks at the performance implications of SSD-based arrays in high-end transactional applications. I think you’ll find it interesting and I welcome any relevant comments you care to make.

Sincerely,

Robin Harris
StorageMojo.com