ORIGINALLY POSTED 27th February 2018
This week, while recovering from the crazy busy week over in Vegas for IBM Think, I was sent an email by our competitive analysis team which reminded me of why I started this blog back in 2007.
In those days, there were only a few of us, Tony Pearson (IBM), Barry Burke (EMC), Hu Yoshida (HDS), Chuck Hollis (EMC) and of course, Steering Wheel Cam of America, 3Par Marc Farley. (Last I knew, with HDS). We’d all bicker and banter between us, particularly when someone made a big blunder, or missed the point, notice Chuck and BarryB – and their old blogs, seem to have vanished!
Well here we are in 2018, and it seems that our friends over at Pure Storage could do with either educating their CEO, or maybe they need to invest in their own competitive analysis team, heaven knows they should have some spare cash since we helped them by debugging their SCSI front end implementation no so long again.. Ordered Test Unit Ready anyone…
Anyway, the article in question is over on Search Storage, where they interviewed Charlie Giancarlo, their newish CEO and he makes one interesting, and one plain incorrect statement.
First, as most of you know, back on Feb 20th, IBM announced support for NVMe-oF (over Fabrics) on our FlashSystem 900 Gen2 product line, which first implements this on our Infini-band interface ports. We’ve also issued statements of direction that the majority of the rest of our block storage products (SVC, Storwize, V9000, A9000) are all ‘NVMe-oF’ ready – that is the hardware can support it, and watch this space for a software only update to unleash it.
So its quite amusing to see Mr Giancarlo state :
“I’m very certain we’ll be the first [storage array vendor] with NVMe over Fabrics, but we don’t have a date on that yet.”
Err, nope, you are not certain at all, infact you are wrong. And I wonder how close you really are, as you don’t even have a date yet. Not only that, but this was covered quite extensively that the time : The Register and 451 Group just to mention a couple.
“The headline is that 20% of our sales last year were for NVMe.”
My second point is regarding the 20% of what they sold being their NVMe based Flash array. This I find curious. Let me explain. back in April they launched their update to the Pure storage array to include NVMe. Given that the majority of the IT budget spends are in 3Q and 4Q of a year, that would say you’d expect the majority of their annual sales to be after the launch. Yet only 20% of the boxes went out as NVMe? We tend to find with Storwize and FlashSystem, as soon as we launch the new generation, almost all new purchases are of that new generation. So thats why I find it curious that only 20% was NVMe, so 80% still the old box… maybe people are scared of picking up such a new technology. Particularly when its only using single ported drives. Thats like taking a desktop SATA HDD and putting it in your high end controller. At least NL-SAS is dual ported.
So while IBM hasn’t made any public statements about internal NVMe device support for our block storage, you can pretty much assume that we’d want dual ported devices to ensure when a single path or controller fails, we don’t have to ‘take control’ of the device in an active/passive way – it will be truly active/active.
That also brings me towards the end, one thing Pure seem to be good at is convincing the C level executives that they have a great product, while I have seen the technical teams that have to then implement it are less than enamored with the reality. I mean come on, active/passive controllers, how very 1990’s (Data General) – what became EMC Clariion – and even EMC gave up on that as a bad idea. Maybe that’s why Pure need so many cores to run their latest box, because half of them are sat there unused most of the time.