Pure Storage pushes all-flash array purification Pure Storage is announcing a fresh all-flash array family with capacity and performance boosts, updated cloud and analytics-based support, as well as enhancing its upgradability business model. FlashArray//m is the hardware, Pure1 the cloud and analytics support facility, and Evergreen the upgraded upgradability scheme. Given …
So basically they have two 2.5" SSDs together in each slot over a SATA/SAS interface? Thats not custom designed, just putting 2.5" SSDs plugged into each slot.
Also when you add another shelf then the first array becomes a top of rack controller with the same CPU and DRAM constraints as before? I guess this is why the max limit is 4 shelves per controller without scale-out.
Apart from moving to 2.5" SSDs and the controller taking some SSDs, I can't work out whats different about this solution? The software seems the same with no new features.
The density, performance and software feature is still lagging behind some of the other vendors such as IBM, Violin Memory and HP, not exactly industry changing stuff here.
I am confused . com
I like the bezel though, that looks cool :-)
Interesting to see how the OEMs are playing with PCIe based NVMe flash to get that last bit of IOPs and latency out of the interconnect to the flash.
SAS/SATA interfaces were never designed with flash in mind and fitting silicon to what was intended as an interface for rotational media is not the best use of that silicon resource.
I would surmise that in a few years all SANs (virtual and physical) will leverage to some degree or another the PCIe as the preferred interconnect to the flash media.
I love how these raunchy startups are pushing the technology in new directions and forcing the large OEMs to scratch their heads and announce 'why didn't we think of that'.
Rock on Pure!
So Pure called this the "beauty"? Sure, it's lipstick on a pig. It doesn't change the dynamics of their portfolio, which by now has seen more than 6 models. OK, it's cute in terms of design but under the hood the only thing new is that they do finally have NV-RAM, instead of slapping that term on SSDs, with no consequence on their performance nor their lack of scale-out (and interestingly they abandoned Infiniband, so a distributed architecture seems less likely). I also can't see how you would non-disruptively migrate from the FA-4xx series to this //m series, since the hardware is incompatible, which makes me think that, errr, you don't. But yes, the hardware is cute.
This post has been deleted by its author
OK so there are 20 SSD bays each able to house 2 SSDs with a max capacity of 1TB per SSD. Wouldn't that equal a 40TB design?
How does the Pure Storage website, video and PR claim that a FlashArray//m can pack "120 terabytes in a 3U chassis" or "40TB per U" ???
I've never understood the issue with this - it's the same with our Data Domain - my device deduplicates data and so this is the estimate based on the data set you've given us. Yet everyone loses their mind every time they see this raw type of calculation....unless i'm stumbling upon competitors just trolling. I dunno.
Not looking into our tier 1 for a few years but the thing I like, and didn't know, was the flat maintenance forever - year 1 cost of maintenance is the same as year 10/15/20 whatever. Is this a new thing or part of the announcement?
Seems like old wine in a new bottle...yawn...
This just means that Pure's margins will move up a few more notches. As it is with a 6:1 data reduction ratio, they are priced exorbitantly compared to other arrays. Most hardware is "purpose-built". Not sure what purpose the 3U serves for the customer.
Honestly, the whole Pure story seems to me like a pump and dump. The product does not have a great architecture.
I think:
* A bunch of people join Pure to get a heap of options
* Pump the product up to be the second coming
* Wait for IPO
* .....
* .....
* .... more waiting
* IPO!!!!!
* Head for the exits
Pure is heading for a nasty crash. Nothing they can do about it now. They needed to get out into the stratosphere when Nimble did their IPO, but missed the launch date. The marketing hot air is losing steam as their differentiation has already dissipated. They have nothing to delay the hard landing but the fading supply of ballast from being "an early flash vendor." The big boys, EMC/XtremIO and HP/3PAR All-Flash, have already surpassed them before the ink on their S-1 has had time to dry. According to VARs I've spoken to, Pure's desperate strategy of over-distribution in an attempt to keep the growth going have annoyed the very Channel Partners that invested in them early, and many of those VARs are now turning away.
There's only one question left. Will it be a slow crash landing as the air leaks out, or will some poor bunch of mutual funds and other investors get suckered into an inflated IPO and then be bird's eye witnesses to a mid-air explosion of Hindenburg-like proportions?
Flame retardant anyone? The Welllington's, Fidelity's and other funds at IPO may need it since its likely to be Pure Hell.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
