EMC is using Samsung NAND in its enterprise flash drives, marking its first second source for flash drives alongside STEC. Traditionally storage array vendors like EMC like to have second sources for technology such as disk drives, hating being dependent on a single vendor. Thus they can use disk drives from Hitachi GST, Seagate …
"Samsung reckons its enterprise SSD prospects are good, quoting Gartner numbers claiming worldwide SSD units and revenues for enterprise application will grow from 324,000 units and U$485m in 2009 to 6.3 million units and U$3.6bn in 2014."
Which means lowering the average price per drive from $1400+ to a more palatable $571
Burn, wallet, burn ...
Pffft always with the sequential specs
As EMC is well aware, yet never mentions unless backed into a corner... Not all I/O is sequential, and it's the random I/O performance of solid state storage that is really the game changer. Maybe by going solid, EMC will finally be able to deliver decent performance to those of us who deal with random I/O and have found it better from a performance standpoint to stay local.
Of course... this also raises a question: since we're not dealing with platters anymore do existing storage architectures (2.5 and 3.5 drives, SATA, SAS, Read/Write caching... even Raid and filesystem design) really make sense anymore? The next question is, with platters and their comparatively poor performance out of the picture, how much of a game changer will that be for the rest of the infrastructure game - server architecture in particular?
To me at least, reading about EMC building 100-200 GB 3.5 inch flash drives - when they're in the business of selling, at least to their really large clients, petabyte storage infrastructures - just seems a little uninspired.
Of course, in their defense, solid state is relatively new to the game when it comes to mass enterprise storage so maybe I'm just being impatient... and their business is built around hundreds of thousands of 3.5 drives after all. Virtualized mass storage (vs. local or dedicated spindles on SAN) with solid state, even just plugged into existing SAN architecture *should* be more attractive for us random I/O folks - but until I see it deliver on its performance promises, whatever they actually are, I'm not holding my breath.