The ‘Myth’ of the Mainframe Cloud

December 5, 2013

OMGOMGFLUFFYPINKRAINBOWUNICORN!!! But is this more likely to exist than a mainframe cloud?

Recently I read an interesting piece by Quentin Hardy (@qhardy), the Deputy Tech Editor and The New York Times on IBM’s Big Plans for Cloud Computing. Featuring Lance Crosby (@lavosby), CEO and Founder of SoftLayer (recently acquired by IBM), it raised the always-expected but never-delivered notion of an IBM mainframe cloud.

Not surprisingly when such topics are mentioned, it also drew derision from some quarters who think the idea of a mainframe cloud is a fanciful myth.

This time around, the comment came from Dell’s Bernard Golden (@bernardgolden):

I quote this not to pick on Bernard – he is a smart guy whose opinions I frequently respect and learn from – but rather just to show that this is a very common view even among the most educated pundits in the market.

I am not going to talk about the customers I know that have built, or are building, private clouds on mainframe IFLs – the Integrated Facility for Linux processors that are a core part of the System z architecture. Mainframe cloud has been on enterprise plans for years, so this is just old news.

And yes, they are ‘true cloud’ (eyeroll), with on-demand self-service; broad network access; resource pooling; rapid elasticity; and as a measured service. Not that definitional exactitude is important to the business value their mainframe clouds are delivering.

Even beyond this current reality, there is any amount of evidence to suggest IBM in particular has a substantial opportunity in providing a public mainframe cloud – IaaS or even PaaS. Despite all odds (and predictions) mainframe is – depending on the quarter – still a growth market, albeit at mostly marginal rates (1-2% p.a). Yet on the same day Hardy posted his article, Matt Eastwood of IDC (@matteastwood) reported:

Remember that this is just for new MIPS. And for hardware sales alone. If this rate was sustained over the whole year (which is very unlikely), then just substituting cloud for new mainframe purchases makes an annual market of $3.6bn.  Much of this growth is likely the product of inertia. Many organizations with mainframes would much prefer to use open commodity PAYG platforms. As Mark Griffin (@grifmon), Enterprise Integration Architect at Genworth Financial, noted:

Others like Mark will go as far as to eliminate their mainframes entirely, and many more would like to:

This seems logical, and smart for many CIOs, which is why some (not all) pundits predict mainframe markets will shrink over the mid-term. This seems reasonable too, even though there are also organizations that are continuing to actively invest in their mainframe systems, and for very good reasons – like scalability, throughput, and availability.

However, few mainframe CIOs can just dump their System z investments onto commodity cloud, even if they wanted to. A large public mainframe cloud doesn’t really exist; it is tough to justify investment in porting, let alone rewriting entire applications; and mainframe applications are so complex, ingrained, and mission-critical that the risk and effort is huge. Meanwhile, as Mark notes, there is a day-to-day IT business to run:

And who has time for that and a mainframe migration project with massive risk and little payoff? So the inertia is with maintaining current services, growing mainframe as little as possible, while avoiding mainframe for new applications. This inertia both helps the mainframe market to grow at the margins, and inhibits enterprises from migrating to commodity cloud services. Yet, the growing market for new mainframe MIPS is just one opportunity for mainframe cloud. There is also the market for ‘lift-and-shift’ onto a fully-compatible mainframe cloud, to take those ‘lights-on’ processors off the CIO’s books. Not to mention the non-critical dev/test mainframes that could be easily replaced with a public mainframe cloud. Objective commentators who are learned about these issues – someone like Gartner’s Lydia Leong (@cloudpundit) – certainly get this:

Even Simon Wardley (@swardley), often curmudgeonly on how disruption leaves stale old businesses behind, gets this:

This is not even considering the opportunity for ‘truly’ innovative mainframe cloud services.

No surprise then, if IBM/Softlayer end up doing exactly this. IBM/Softlayer exec Crosby was very explicit that IBM has big (iron) plans for a mainframe cloud:

Mr. Crosby said, IBM is “absolutely” looking to sell its big mainframe computing capabilities as a cloud-based service.

Of course, many big questions still remain – like when, what, and how much. Everyone has expected IBM to do this for some time, but it is surely (as Lydia Leong put it), a non-trivial exercise to implement. So when will it be available? How much functionality will it have? And what will it cost?

I certainly expect IBM to follow through on this eventually. It really seems just a matter of time.  I will be excited to see it if or when it does come, as a mainframe cloud will usher in a world of opportunity for mainframe enterprises. Meanwhile I will continue to see private mainframe clouds delivering real business value, and proving true the myth of the mainframe cloud.


(image via Spin-Art at DeviantArt)

Tags: , , , , , ,

9 Responses to The ‘Myth’ of the Mainframe Cloud

  1. December 30, 2013 at 12:41

    Thanks, Andi, for exploring the tension many IT teams are facing as they try to maintain day-to-day operation and consider migrating to a cloud mainframe. I have a question for you about your last sentence “Meanwhile I will continue to see private mainframe clouds delivering real business value, and proving true the myth of the mainframe cloud.” So the private cloud demonstrates real value and further dispatches the concept of a cloud mainframe into the ether? It sounds like you might believe that the rising popularity of private mainframes may obviate the need for a cloud mainframe?

    Thanks in advance for your time.


    • December 30, 2013 at 14:00

      Hi Jennifer, thanks for reading, and for your comment. That last sentence isn’t really clear, is it? Apologies, I actually meant a couple of things there.

      To clarify, I believe private cloud (mainframe or x86) is not going away, and I already see large orgs using owned mainframes for their own private clouds. But there is also demand for a public cloud built based on a mainframe platform, for migrations and new development on on-core mainframe applications. These will continue to co-exist, I think.

      I do agree the emergence of a public option for mainframe applications will reduce demand for owned mainframes eventually, but it will not eliminate it. There will continue to be a need for both private mainframes (for traditional and private cloud use cases) alongside public mainframes (for cloud-migrant and cloud-native use cases).

      This is how it seems with most new IT developments. We do horrible job of moving off the old; we just add the new alongside it!


      • December 30, 2013 at 16:29

        Thanks, Andi, for providing additional clarification! I agree with you that there will continue to be a need for both. What will be interesting to see develop are the types of tool that will emerge that allow IT teams to automate when one is used over the other or that allows switching between the two depending on different capacity thresholds or security issues.

        Have a great start to the new year!

  2. December 9, 2013 at 10:18

    There’s a line of scripture from the Book of Ecclesiastes (chap 1:9)
    “The thing that hath been, it is that which shall be; and that which is done
    is that which shall be done: and there is no new thing under the sun.”
    A little history as I have been in the IT field over 31 years now (back then it was called DP or Data Processing) In the late 60’s on thru the 70’s there were business entities called
    Service Bureaus. They offered IT services to those companies that were large enough to require automated processing of their business information, yet were not large enough to afford their own mainframe.
    Some relief was available thru the first Unix and Windows servers in the 80’s.
    So here some 40 years later we have this supposed new technology of “The Cloud”.
    It’s really a new twist on an old idea.
    I still find it curious that this so-called “mainframe vs non-mainframe” debate still
    continues. What’s really at play is looking at a business’s IT needs and sizing a system to meet those needs with the flexibility to expand or contract IT services as the business needs change. The cloud concept offers this flexibility.
    Just as in nature there are little clouds and big clouds, it is true of so-called cloud-computing. Large concerns with the need to process millions of transactions daily
    will require the services of “Large Cloud”, a grouping of small businesses with only say
    a transactional demand ranging from a few hundred thru a few thousand transactions per day can operate on a smaller cloud.
    The real mistake being made is trying to come up with Unix/Linux/Wintel based servers that
    mimic the virtual system that IBM has had in play for over 50 years now, by cobbling together
    rows upon rows of blade servers running VMware and consuming tens of thousands of kilowatts
    in a vain attempt to offer the same processing power of a zEnterprise System.
    I’ve worked in shops where these Unix/Linux/Wintel server farms consumed far more energy than a fully tricked out z10. People complain about software licensing and pricing; last time I checked Oracle and VMware ain’t all that cheap.

    Yes there will be private clouds and public clouds, large clouds and small.
    When cloud computing really takes off and I don’t doubt that it will, all of these hardware/software manufacturers will find themselves with shrinking market share as more companies dump their IT departments in favor of cloud services when it’s in the interest of their bottom lines to do so.

    • December 12, 2013 at 21:57

      Guy, thanks for reading and for a great comment.

      I agree more or less with the history. Hang around IT long enough like you and I have, and you see the patterns, don’t you? Everything old is new again, eventually.

      I do think there are unique advantages to cloud – and indeed, every iteration of core ideas (like OOP –> SOA –> Web Services –> APIs) brings new specific benefits.

      But I am 100% with you on the importance of looking at the value and use case, then selecting the underlying platform to achieve the goals. Mainframes do have specific and unique benefits, but do not (and should not) run every workload. For that matter, same could be said of x86.

      Ultimately it has to be a business decision on what applications and services – size, scale included – which dictate the platform, and the provider. Not the other way around. And the best option for the bottom line will almost always win.


  3. December 9, 2013 at 02:35

    Firstly let me congratulate the author on a very measured and balanced post, I mention this first up because in the religious fervour of platform selection such a balanced summary is not the norm. Let me address some of the points raised. IBM and others such as CA see the mainframe as a platform for both public and private cloud services, and yes I mean true cloud measured against the definition given in the blog post. The number of MIPS delivered as Linux MIPS is increasing year on year and after 13-years of support from the platform for the Open-Source OS it is now reaching a level of adoption and maturity that is an interesting dynamic in the wider marketplace. I put this down to the speed and architecture of the chipset delivered with the z196 and subsequent EC12 driving consolidation ratios that make the economics make sense Vs x86. On the point around IBM offering Mainframe (z/OS) as a cloud we have been doing this for some time with the traditional bureau type services and expect this to evolve with the SoftLayer acquisition. If people want to enter a more detailed discussion please get in touch @StevenDickens3 or follow the industry discussion by using #MainframeDebate

    • December 12, 2013 at 22:01

      Thanks Steven, really appreciated your comments here and connecting with you on Twitter. Thanks especially for the specificity around the mainframe growth, benefits, and cloud options – great to hear more about the evolution around the Softlayer capability especially.

      I know we will both be watching this space. As two organizations fully committed to the mainframe, I know we will both be watching this space!