Last year StorageMojo interviewed Violin Memory CEO Don Basile. He noted that as flash features sizes shrank, NAND would get slower as well as reducing endurance.
Intellectually that is correct, but it isn’t an easy concept. Flash stores a physical thing – electrons – and smaller features means fewer electrons, lower endurance and more difficult reads. But accustomed to Moore’s Law, it was hard to imagine how it would play out.
Fast forward 18 months. Violin has migrated from 32nm flash to 19nm. They tell me their new array is denser, faster and more power efficient despite slower flash – with the same number of chips.
Twice the performance with 3x the drives? Maybe there is a problem as flash gets slower.
Here’s a Storagemojo Video White Paper that lays out the issues in less than 5 minutes.
The StorageMojo take
Storagemojo suggested last year that SSDs were not optimal for storage arrays. Not everyone agreed.
Proponents argued that the SSDs massive volumes, low costs and architectural simplicity would inevitably overwhelm whatever performance advantages designers could achieve through a clean-sheet non-SSD architecture. Some went so far as to assure me privately that Violin could never migrate to denser flash while SSD-based designs would automatically upgrade as the SSDs did. Today Violin has migrated, SSDs are denser, and it appears SSD arrays may also be slower.
It will be interesting to see how other SSD-based arrays fare in the months ahead. StorageMojo will be watching.
Courteous comments welcome, of course. Happy to hear other views on the SSD array performance issue.
Disclosure: Violin paid me for this and previous video white papers, but the facts speak for themselves. I’ve been a fan of Violin’s architecture since they briefed me in 2008.