HP 3PAR has picked Hitachi GST flash drives to speed array data access. HP 3PAR F-Class, T-Class, and V-Class (P10000) can ship with a 3.5-inch version of Hitachi GST's SSD400S, the 2.5-inch solid state drive (SSD) co-developed with Intel and using 34nm single level cell flash. It was first announced in November 2010 but OEM SSD …
flash isn't new for 3PAR
You write the article as if this is the first flash drives 3PAR has used. They have had flash for at least a year if not longer and were originally using STEC (if memory serves).
The original drives were judged by some to be too small (I think they were 50GB). In the grand scheme of things I think it was good for 3PAR to go to "slower" flash (assuming the cost works out) because of the sheer number of IOPS can overwhelm most controllers (even 3PAR) in dense configurations.
If your talking about 30-40k random read IOPS in an enclosure that can support up to 16 drives(F-class) or 40-drives(S/T/V class) your talking anywhere from ~480k IOPS in a single F-class shelf (not that they would allow this configuration) to 2.4 million IOPS in a S/T/V-class shelf(again not that they would allow this config) both of which FAR exceeds the maximum IOPS of the entire system as a whole.
The original SSDs if I recall right 3PAR limited you to 4 SSDs per shelf(I think 8 for the S/T/V class to maintain magazine level availability) or something like that, because more than that could overload the upstream controller ports.
As a potential customer (I don't have 3PAR stuff at my current gig at the moment but would love to have it) I rather have more(larger size), slower flash than less faster flash (costs being equal).
What is it with pushing the latest Anobit product at every opportunity?
Wow look, when you bung a 4Gbit interface on a drive you are amazingly limited to 4Gbit throughput. Who wudda thunk it?
All this faster SAS/SATA is fine...
...but what if you need block-level sharing to more than 2 machines from the same storage device?
Try to compare apples to apples for once, a bus does a different job a city car, you can play ones figures off against the others all day but it's still a poor comparison.
That said I'll accept that given every other part of the V-series is 8Gbps they should be connecting them to those boxes at that speed but the older T and F series are 4Gbps throughout.
I don't get it.
How can 2 multi-level cell (MLC) flash be inherently slower than SLC flash but provide almost 2X random performance and prit near the same sequential performance?
Doesn't sound inherently slower to me!