That architecture is already here - designed from the ground up by Violin Memory. Take a look at vmem.com Hot swappable, NAS, software optimized down to each strip of NAND flash. Lowest latency of any array.
Posts by bigdata
9 publicly visible posts • joined 5 Mar 2013
How to design a storage array: NOT LIKE THAT, buddy
Fusion-io opens trenchcoat, flashes cheap PCIe cards
Then there were 3: Micron slowly 'wipes out' NAND flash rivals
How Fusion-io redlined its PCIe flash motor to hit 9.6 MEEELION IOPS
Rackspace: Why we're designing our own cloud servers

Rackspace is behind the curve if they're still using spinning discs. Having Quanta build them saves them 15%, which is no savings since there's no support.
This is the last leg of spinning discs. CIOs aren't knowledgeable enough about flash storage so they try to cut box costs by buying from China. OCP is hardware communism, so everyone gets the same lousy generic performance. Performance needs tweaking. You don;t get performance by buying cheaper boxes.
I don't see how getting rid of a few unnecessary ports speeds up IO, reduce latency, and cut power usage and footprint.
Dell's having a fiddle with Violin Memory - but will it flash the cash?

A bit of a waste for Violin to go to Dell
But frankly, what's so hard about what Nimble, Whiptail, Skera of the world do? They're somewhere in between fast enough and bulk storage. Not much intelligence or IP there.
I think Violin is positioned well for a bigger acquisition, with both SAS and NAS products, great IP and management. If they have a good IPO and good sales, it would almost be a shame for them to get bought out before the fun begins in a year or two, because they have such a lead in the space.
I think this is a hockey stick growth story.