back to article Pure: We've created the Everlasting Gobstopper of Storage – 'Forever Flash'

In recent months, as ever, we've seen a lot of "refreshments" on the storage scene. Violin Memory has refreshed its hardware and software. SolidFire and EMC XtremIO have refreshed their software. NetApp has introduced FlashRay, and Tegile and Skyera have introduced new all-flash arrays. And Pure Storage? It seems it has …

  1. The Islander
    Big Brother

    Intriguing news item but colour me cynical, I have to wonder how “perpetual” storage will pan out in the real world. Is this perchance a marketing spin?

    However, it is interesting to reflect on this from a personal end-user’s long term perspective, especially in comparison to storage of non-electronic data, e.g. in books, on stone, etc. The personal user is increasingly bound to the tyranny of technology advance and challenged to establish a lifelong – in human terms – storage medium for electronic / digital data. As I see it, users can:

    * store on own medium & avoid upgrades, hoping the medium and/or access device will remain usable over coming decades

    * store on own medium but refresh this and/or the access device over the decades

    * store on vendor / cloudy service, hoping such service will not go titsup before an opportunity to retrieve & transfer all data to another medium / service.

    Assessing any of these options – or others you care to mention - involves the usual array of factors - risk, cost, effort, etc. But do personal end-users really do that sort of thing?

    I get a sinking feeling that the world is moving to the third paradigm. If virtualised & software defined storage continue unabated, ordinary users may yet truly become vassals to the emerging digital monarchs.

    1. Anonymous Coward
      Anonymous Coward

      Rolling upgrade

      I've been in a rolling upgrade since 1979 with the oldest file dating back to 1975. First started with tape, naturally, now up to 25 GB raw disk supporting 5.5 GB stored. My biggest challenge isn't media or site recovery from disasters, it's bit rot. Whatever the cause, natural fungus to programming errors that's my pain point.

  2. Yet Another Anonymous coward Silver badge

    It's called leasing

    All they are saying is that pay today's prices and as the flash wears out in 5years they will replace it with the vastly cheaper/GB flash available then.

    So it's like an offer in 1980 to sell you 1Mips of CPU for $100,000 FOREVER.

    1. Skymonrie
      Thumb Up

      Re: It's called leasing

      Nail, meet head.

      Do your eyes also come with x-ray vision?

    2. MityDK

      Re: It's called leasing

      Except it's not just the flash, it's the controller also. So it's not 1Mips for 100K forever.

      CPU performance increases dramatically every upgrade, and maintenance costs are separate from hardware acquisition costs. So not like a regular lease exactly.

  3. NotSoCloudy

    Economically Viable?

    How will this work commerically for them?

    Replacing controllers with the latest and greatest every 3 years, and replacing disks after 5-10years of runtime has got to carry some serious costs.

    Is the 3yr support fee changing?

    1. Anonymous Coward
      Anonymous Coward

      Beyond misleading

      The math works out just fine when you realize:

      1. The flash price is dropping 30% per year (and will likely only need to be replaced after 5-10 years)

      2. You are paying ~20%/year in maintenance on the flash (plus similar on the controllers)

      The long and short of it is... you are easily paying for the replacement in your maintenance costs.

  4. Anonymous Coward
    Anonymous Coward

    Cue applause for the marketing dept at Pure.

    Real controller costs = 10k per controller (20k total, but likely a LOT less)

    Real disk costs at volume = Pennies considering Pure uses only MLC SSD's that come in the commodity SAS attached form factor.

    Also take into account that even MLC drives can tolerate 3 full program and erase cycles per day for 5 years... you can see how even commodity MLC drives have a very robust life.

    Net-Net1

    $ Cost to Pure... a rounding error.

    Marketing effect make for nice headlines.

    Disclosure - I work for a competitor that offers the same replace after wearout program as part of the basic maintenance.

    1. Pascal Monett Silver badge

      Re: I work for a competitor

      Then why isn't your company trumpeting its own products too ?

      Your competitor is making itself known, and in a rather clever manner. Dissing that is not very productive.

      If your company does the same thing then it should start yelling it off the rooftops as well, or else you'll be looking for a new job when Pure has taken the market.

  5. Jan 0 Silver badge
    Stop

    M&S?!

    "maintenance and service (M&S) pricing" - Surely that one's already taken?

    Some abbreviations are immutable, AT&T, LHC, CCCP, etc.

    These Pure Muppets (M is for marketing*) need to think internationally before committing a faux pas.

    * I've no problem with Pure's technology.

    1. The First Dave

      Re: M&S?!

      Surely they should use S&M instead?

  6. Henry Wertz 1 Gold badge

    "Intriguing news item but colour me cynical, I have to wonder how “perpetual” storage will pan out in the real world. Is this perchance a marketing spin?"

    Some companies prefer the predictable costs of an all-inclusive M&S (maintenance and support) contract over the probably lower overall costs but more unpredictable of paying for replacement kit all at once. They could make plenty of margin (profit) and still get plenty of customers. Flash prices are dropping, controller prices will be stable or (most likely) drop, and if the customer needs a higher performance controller, and more storage, I'm sure they will pay more for M&S on that next contract.

    Openstack's storage (like ZFS, and some other cluster or high reliability storage systems), lets you add devices at will (to add more space), remove devices, and offline bad devices, and have the storage be spread out over whatever kind of network of computers you've got (obviously the faster the better.) Using it as a basis for a flash storage system should make maintenance pretty easy (it could either offline flash automaticcally as it approaches the wear limit, using new flash they periodically put in; or they may remove them manually from the pool and add new flash to replace them.) The actual maintenance of an Openstack system is not too difficult, it'd be nice for a field tech to work with I think.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon