This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.
What We Can Learn About Tomorrow’s Tech From Today’s Mainframes
We’ve all read the stories; the mainframe is doomed. FedEx has announced its sunset of the final 20% of its mainframes by 2024, and surely other companies will follow suit. The data center is doomed too. “Cloud-first” means that every company is going to move their applications to the cloud. It makes great copy. Some IT professionals read it because it means opportunity for them, others because it’s an outcome they fear. But is the idea that mainframes are gone hurting enterprises? That seems to be the case.
Always start with facts. It’s certain that the cloud is driving application development trends and that most new written applications is going into the cloud, not the data center. It’s also true that cloud spending is growing faster than data center spending. Insiders tell me that IBM’s mainframes are sold almost exclusively to current mainframe users, so the market’s not expanding. Most enterprises tell me that they’re pretty sure that they won’t be opening new data centers. Cloud providers tell me that everything is moving to the cloud, that every enterprise agrees, and wouldn’t they know best?
It depends on what you mean by “everything.” It depends on what “every enterprise” really means.
Let’s start by admitting that people have been saying everything is moving to the cloud since the dawn of the cloud era, about twenty years ago. “Everything” hasn’t moved yet, and we’re still hearing it will. That should make us suspicious of that sweeping prediction, to say the least.
Now, let’s link the application development point into our story. What’s been happening over those two decades is that enterprises have realized that the cloud, combined with the Internet, is the perfect way to accomplish the almost-mystical goal of application modernization. If you glue a cloud-hosted front-end software component to a legacy application, you can make that software component accessible to customers, partners, and even workers using a browser or a smartphone. You contain the cost of current technology because no change to that legacy application is needed, and it gets a new and modern look. What’s this model called? The “hybrid cloud.”
In the old days (yes, twenty years ago is the “old days” in tech), people thought that a hybrid cloud was a combination of public cloud services and private cloud. (Of course, what the heck is a “private cloud?”) It turns out that 1). enterprises want to avoid massive changes to legacy applications to host them differently, 2). few enterprises require a “private cloud” as part of a hybrid, 3). most enterprises think that “private cloud” technology means virtualization in any form.
So, can you virtualize a mainframe? The first broad commercial application of virtualization ran on mainframes — IBM’s VM operating system goes back 50 years. IBM’s current Z-series mainframes have used virtualization from the start in the early 2000s. Shall we say that everything new is old again? Or just that mainframes can be an essential piece of that hybrid cloud?
Maybe you think I’m viewing an old-fashioned subset of the enterprise space, where CIOs sit on tall stools and wear green eye shades. Perhaps, but (apparently) an important subset. In the most recent quarter, guess who beat Wall Street revenue expectations? IBM. Guess what IBM’s big pitch to prospects was? Hybrid cloud. Guess who sells mainframes? IBM. Guess what was hot in IBM’s hardware portfolio in the most recent quarter, with sales up almost 70%? The new z16 mainframe. A lot of server vendors would kill for growth a tenth of that.
From all of this, you might think I’m arguing that mainframes are not doomed, and that’s only semi-true. In the long run, if we consider a “mainframe” to be a big monolithic device rather than a rack of things, then the mainframe is doomed. In the near term the mainframe is not doomed. It’s gradually evolving into a newer vision of the mainframe as a partner in hybrid cloud rather than dying. In the process, the mainframe has already redefined “hybrid cloud,” and it’s going to redefine it even more in the future.
IBM’s view of hybrid cloud is mainframe-accommodating, not mainframe-specific. A hybrid cloud is an application model where public cloud services act as an on-ramp to legacy applications running on whatever they run on. No evolution of data center resources is mandatory. The data center will evolve as it always has, under the positive pressure of application and business changes, and the inhibitors of capital depreciation, challenges in changing application software, and IT operations inertia. When, over time, the value of the new overcomes the difficulties of abandoning the old, things will change.
Its clear mainframes will eventually be superseded by this value of the new exceeding the difficulties of the old, but the data center? From the very first days of public cloud services, enterprises have been consistent in telling me they did not believe that the public cloud could offer them data security. They still say that. Yes, a few are talking about moving everything to the cloud, but that’s a minority view. We use the term “data center” to reflect a central computing point, but it’s really about data security and access efficiency. We put data centers on-premises because data gets used a lot there. When I ask enterprises themselves how many enterprises will go totally to the cloud, the answer hovers in the 25% range.
In any event, we shouldn’t define cloud success by what gets moved to the cloud or how we hybridize the cloud with the data center. Past IT transformations like minicomputers and personal computers transformed things by doing what wasn’t done before, not by moving or tweaking current applications. The “move to the cloud” hype has covered up that truth, has kept us—even IBM— from asking what the cloud could enable that has never been done at all. The question isn’t whether mainframes are doomed, which is irrelevant. It’s not whether the data center is doomed; that’s been answered in the negative. The questions are how we will use IT in the future? What vendors own the future. Forget the hype and start working on finding answers to those questions.