back to article Reg man has the cure for IBM storage: Just swallow 10 firms

When you are walking on the Moon, far from home and Earth's tethering gravity then giant steps are what you take. When a product set becomes low or no-growth, then its tethering product gravity effect is weakened and you need to take giant steps to rebuild it ... which brings us to IBM. Witness its results. IBM's old …

  1. danXtrate

    Dear Reg Man,

    I think you need a little more insight in the storage ranges of EMC and IBM.

    For once, IBM has quite a great combo between SVC/Storwize/FlashSystem and Tivoli Storage management products and Cloud Manager.

    You have SmartCloud Virtual Storage Center as a Software Defined Storage management gateway with which you can easily integrate whatever storage you might consider.

    Storage Virtualization is years in front of ANY other competitor. EMC basically give VPLEX away with midrange VNX arrays just to be able to compete with Storwize v5000 and v7000.

    As for Flash, why bother with hybrid arrays (which you can have just by adding SSD's to Storwize) when you can virtualize FlashSystem with storwize v7000 (Gen2 especially) and San Volume Controller?

    Where IBM is truly lacking right now are the unified versions of Storwize v3700 and v5000 arrays, EMC is quite a few steps ahead in the game in this area.

    SONAS and Isilon are quite evenly matched, but again at the high end. Isilon can be better at the lower end of the scale out NAS market.

    If you look at the IBM Storage division results patterns over the last couple of years you could realize that the major market losses coincide with the scaling down and eventually dropping out of the OEM partnership with Netapp.

    IBM was developing their storage portfolio with LSI and since they were bought by Netapp most of IBMs entry and midrange portfolio was in the hands of a major competitor. All of the DS and N Series ranges were phased out and replaced by Storwize. It was quite a hit, but I see IBM recovering from it.

    I'd really love to see Storwize and FlashSystem being sold to Lenovo, IBM has a really hard time marketing their products to end customers while EMC has some great marketing programs. We'll see what next year will bring, Lenovo seems to be in a buying mood.

    1. Anonymous Coward
      Anonymous Coward

      > most of IBMs entry and midrange portfolio was in the hands of a major competitor.

      You do realise that IBM resell NetApp via GTS? Just because STG has stopped the OEM does not mean that selling has stopped.

      1. Anonymous Coward
        Anonymous Coward

        Why would anyone buy NetApp through GTS, at a hefty markup, when they can buy direct from NetApp, who have quite a competent sales force. The only place IBM really sold N-Series (NetApp filers) was where a customer wanted an "all blue" solution. IBM adds very little value, and plenty of cost.

        1. rvandolson

          Discounting? We've always gotten better pricing on N-Series vs. equivalent FAS. Maybe we could have been more thorough on checking more VARs.

          Also, there is value in minimizing the number of vendors you're doing support renewals with if you've got a large footprint (yes, and conversely there can be value in having multiple vendors to play against each other).

      2. danXtrate

        Yes, I do. But you do realize that GTS figures are not taken into consideration when reporting storage numbers, right? GTS is just a large VAR from my point of view, they would resell whatever the customer wants.

        @rvandolson Have you considered a migration to Storwize Unified? I believe you can virtualize older storage with a unified Gen2 box now and with the added compression cards you will bring new life to ageing NSerries boxes. Provided you also use FC of course. How do you addredd the extra bandwidth needs in the Bladecenter chassis? I've had customers migrating to Flex rather than upgrading their current Bladecenter chassises because of complexity of upgrades and cost of investment in an ageing architecture.

        1. rvandolson

          @danXtrate - Yes, definitely interested in Unified, but have shied away thus far more or less waiting for it to mature. At EDGE this year a number of folks I spoke to indicated they still were hitting bumps in the road and felt it didn't match up close enough with FAS. I want to revisit again in 2015 though.

          So far we haven't run into BW issues with the BC chassis -- could be a side effect of our workloads, but we also do dual 16Gb connections from each in-BC switch (two switches) up to our SAN96B's. I think we may switch to using pass-thru instead of an in-BC switch for our next round though. Using Dell BC's, btw.

    2. rvandolson

      Dan -- thought many of the same things as I read the article. No hybrid solution? What do you call the V7000's I have mixed with SSD's and HDD's? I still find them a better fit (and less expensive) than the AFA's out there today where you also have to rely on getting advertised compression and deduplication rates... also it all fits nicely under TPC and I can virtualize my older storage (probably even my N-Series boxes which will be fine for years to come with third party support) behind SVC (standalone or built into the V7000).

      I'm hoping their unified solution continues to mature -- we have lots of N-Series and that disruption has been a messy one for us.

      Count me in as a customer who still likes most of IBM's storage portfolio. The Storwize line is great, and I even prefer LTFS/LTO for cold data archival rather than the latest trends like Glacier (it's less expensive and when someone needs to pull satellite imagery they don't need to wait for it to get pulled down via our internet circuit!)

      With all of that said, we'll likely dip our toe in the hyperconverged space sometime next year. Don't know that it'll be cheaper -- just a different way to get much of the same things done. We're blade center based with FC infrastructure already built out, so costs would seemingly shift to needing more network ports to accommodate the Nutanix-style pizza boxes (maybe there will be more blade-based options soon).

  2. Anonymous Coward
    Anonymous Coward

    IBM Cures

    Ginni does not want to cure this ill sadly. IBM has a lot of very smart people who would love to dominate this market but they cannot develop without funding. It's too late already as the market splits between Flash for the special data and low end disk for everything else. Who can make money selling at $750/GB and dropping for low end? Who can make money competing against all the Flash startups selling at cost or lower? How long before hyperconverged/EVO drains the need for external storage in all but the larger enterprises? Those of us with some gray hair will recall when all storage was internal and only went external when servers couldn't hold enough disk for the data. HC/EVO and Flash/fat drives in servers solves that problem and simplifies life for infrastructure teams. Not saying it's perfect but neither are SANs. Still think IBM should invest in storage?

    1. Speltier

      Re: IBM Cures

      "...but they cannot develop without funding."

      Dropping that 20 EPS target might free up some funding. Future product funding is a major worry for long term customers, who worry about investing in constructs that won't have any legs not due to competition but due to neglect.

    2. danXtrate

      Re: IBM Cures

      Server based flash and spinning drives may serve their purpose, but arrays still serve their purpose. You cannot get the availability and redundancy a serious virtualized environment needs. Software defined everything companies promote that software wil rule everything, but fail to realize that without a serious hardware foundation the entire structure will collapse. And to be honest server based controllers are not nearly as powerfull as array based controllers. You could scale out, but that has it's own performance challenges as you add and add nodes.

      1. indulis

        Re: IBM Cures


        The server based controllers running GPFS Native RAID software in the IBM ESS and GSS storage units are typically getting in the order of 4-7 GBytes/sec per server controller, including lots of advanced functions (disk hospital, declustered RAID, as well as a parallel filesystem layer), so I'd argue with your assertion that array based controllers are better. Scale out with ESS or GSS Elastic Storage (GPFS) based units isn't a problem, look up MIRA at Argonne. 240 GBytes/sec second from a single filesystem with multiple GSS subsystems hooked up together. If our competitors can't use the same systems to get to these performance levels, and can't use the same technology to also support commercial uses (e.g. DB2, Oracle, SAP, MQ, SAS, TSM) uses in a scalable... then please don't make the assumption that IBM is in the same boat!

        [Disclosure: I work for IBM, many years in commercial & technical, big systems]

    3. Yes Me Silver badge

      Old story [was IBM Cures]

      > IBM has a lot of very smart people who would love to dominate this market but they cannot develop without funding.

      Sorry but IBM has messed up this market for years. Killing AFS (twice). Utter waste of resources on DFS. Ignoring NFS. HPSS. StorageTank. Bringing GPFS to the fore ten years too late. Frankly, unless the corporate culture in the storage area (that there must be One True Blue Solution and the rest of the industry has everything wrong) has undergone a dramatic change (other than utter demoralisation which I assume continues) I don't see anything getting better.

    4. El Storage Master

      Re: IBM Cures

      Your hair must be super gray if you think flash (or other high-end) storage is happening at $750/GB! Do you realize the street price of most all-flash arrays are below $10-15/GB (raw, without de-dupe) at this point? Do you also realize how smart of an acquisition TMS was for IBM, and how this proves they are capable of making further great acquisitions? Being less worried about the future of your legacy disk business than every other vendor is, happens to be IBM's biggest strength right now.

      1. Anonymous Coward
        Anonymous Coward

        Re: IBM Cures

        If only AFA's were that low. Recent quote (last 45 days) from Pure (their FA420) compared directly against V7000 came out to ~$14K/TB raw. They assume 6:1 data reduction for typical VMware type workloads so that gets me to about 79TB usable or around $4K/TB. EasyTier based V7000 is still quite a bit less depending on your discounting and mix of spinning disk. Pure showed a 5-year TCO savings against V7000 but most of that was based off of administration time overhead and a presumed big hardware tech refresh needed on the V7000 side at the end of year five.

        I'm sure at some point we'll reach a tipping point for AFA's, but why couldn't I just do it with V7000 exclusively? Maybe they'll add in dedupe at some point in addition to the compression...

  3. ecofeco Silver badge


    They are ossified.

    Of course it's well known I have very little respect for IBM these days, along with thousands of others.

    The only bright spot I see at IBM is their Tivoli family of software. From storage to desktop support tools, it's pretty decent kit. Is there better? Of course, but Tivoli tools are actually not bad at all for general everyday use. These days, that's saying a lot.

    Oh, and the fact that they've switched their employees' desktops to Linux. That really quite surprised and impressed me.

    Other than that, they are a petrified turd.

    1. Anonymous Coward
      Anonymous Coward

      Re: Ossified

      "The fact that they've switched their employees' desktops to Linux..." has one small failing, it isn't a fact.

      Some people with access to key client sensitive data are forced to use Linux but the rest of us are quite free to continue with the latest flavour of Windows, and do.

      1. Anonymous Coward
        Anonymous Coward

        Re: Ossified

        " "The fact that they've switched their employees' desktops to Linux..." has one small failing, it isn't a fact."

        It is a fact that they attempted this. It is also a fact that it did not go well, and they have largely retreated from it. As you say, only a relatively small number of people are forced to use Linux for security purposes, and they generally end up just using it as a VM host for Windows clients to provide an extra security layer.

  4. alwarming
    Paris Hilton

    Can't swallow 10 firms all at once....

    That's what she said.

    Paris, coz she'll handle the staff.

  5. MadMike

    New POWER8 servers:

    IBM has released new POWER8 servers, the largest is P880 having 16-sockets with 16TB RAM. This is quite small, considering that x86 has 8-socket 12TB RAM servers now. Why are there no articles about the new IBMs large POWER8 servers?

    If you need some serious horsepower there is no other alternative than 32/64-socket SPARC servers, sporting 32/64TB RAM.

    1. klaxhu

      Re: New POWER8 servers:

      nobody writes about these things because sparc, risc and power even itanium are dead. all companies who still sell these are either bleeding money in that business unit or planning to move over x86 completely (oracle/sun, hp even on non-stop, etc)

      vmware killed them, even though they were great products. somehow, they didn't stack up to the vmware cost model and the ease of use of the cloud.

      yes, they will still have their market as the mainframe has it's slice, but very tiny.

      nobody cares about speeds and feeds on high end cpus/platforms anymore because they cost so much think about it: it will still work for a while as there are apps designed for these platforms, but every new app now gets written for stacks that are in the cloud. developers and especially new companies don't give a damn about technology, they just want their app to work so they get vm's on aws, azure or just build their own stack on vmware or hyper v. .or kvm

      that is why nobody cares about the mighty CPU's anymore...

      1. Anonymous Coward
        Anonymous Coward

        Re: New POWER8 servers:

        > vmware killed them, even though they were great products. somehow,

        > they didn't stack up to the vmware cost model and the ease of use of the cloud.

        Care to elaborate on this ? Not that I disagree, just that I am unaware of the costs involved.

        1. danXtrate

          Re: New POWER8 servers:

          VMware did not kill them, they have just repositioned. You cannot get the availability, flexibility and uptime of a Power based machine or VM on x86. Not yet at least.

          KVM is supported on Power CPUs, by the way.

          1. Anonymous Coward
            Anonymous Coward

            Re: New POWER8 servers:

            << You cannot get the availability, flexibility and uptime of a Power based machine or VM on x86. >>

            Try telling that to the thousands of companies that have mission critical IT systems running on x86.

            Yeah, you really can - and you have been able to for many years, now. That's the benefit of software defined infrastructure.

  6. Anonymous Coward
    Anonymous Coward

    Time to boost their storage portfolio

    Reg Man,

    They need a solution that will help them step it up in the data management / data protection space in storage. They need to invest in a copy data solution. Rumor was at their Edge event in 2012 they previewed the solution and then dropped it. At the same time they dropped all future development for their dedupe solution... They need to step it up and solve that challenge for their customers.

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

Other stories you might like