Mainframe Modernization, AI, & LinuxONE with Steven Dickens at SHARE Cleveland 2025

We are at the Annual Share Conference here, which is the longest Running tech conference, and I'm catching up with, uh, my old friend and comrade, Mr. Steven Dickens, uh, who straddles the world of Mainframe and Open Systems. Has a lot of experience in both, and I, I like to consider him the mayor of share.

So Steven, welcome. Uh, talk to us a little bit, uh, first off about yourself and your background. Like, like what are you doing here?

So, CEO and Principal Analyst of Hyper Frame Research. We cover everything from Cloud observability, Kubernetes, all the operation stuff regardless of platform. But where I go deep and where some of my technology sort of depth comes was I was a product manager in IBM launched mainframe products.

LinuxONE was my baby. Me and another product manager launched that just 10 years ago. Mm-hmm.

In, on the 17th of August, I've been the chair of the Open Mainframe project. When that project launched, I got the pleasure of working with the Lennox Foundation to launch that project. So broad analyst coverage of the space, but also the ability to dive deep and roll with some of the nerds that are here this week and be able to understand the technology at a, at a pretty low level.

And, and that's valuable both to the mainframe guys and to some of the other vendors that are out there. We work with people like AWS and people like that who are looking potentially to get work out, load off the platform. So I can bring value to both sides of the equation.

Well, and and I think that that may surprise some of the people listening to hear AWS mentioned in a mainframe context. They're here this week. They are, and first time ever.

And we've talked to them as well about, uh, the work that they're doing. Um, so again, if people are just joining us and they're, and maybe they think, oh, mainframe that's dead or that's a dinosaur or whatever, um, what is the current state of affairs like? Like give us the sort of the, the 500 foot view of what is the mainframe today.

So I think we hear a lot about mainframe modernization as a term. It gets banded around, it gets banded around by people who want to get people off the mainframe and it gets banded around by people on the mainframe platform. We wanna modernize applications.

The press gets involved. We hear things like 50-year-old applications. I think where we are today is we're in an inflection point Z 17, the latest mainframe was launched.

It's got an AI accelerator. It's IBM's newest mainframe hardware platform. Yeah.

About every two and a half, three years they launch a new box. We're talking about a five nanometer, um, chip technology runs at over five gigahertz, huge memory footprint. This thing's a beast.

It runs Linux, it runs Red Hat, it runs suse, it runs canonical. It's as mod, you can run Ansible on it, you can run Red Hat OpenShift on it. It's a modern platform both from the speed, it's got a GPU accelerator, sort of DPU accelerator in there called Spire.

If you were to look at it and not call it a mainframe and show it to any hardware techies, they would go. It's a modern platform. And, and one thing I heard you mention, I don't want to overlook that, is that they specifically have an AI accelerator that they're working for it.

Yeah. Yeah. And that's cool spa.

Yeah. And that's gonna be pretty cool. Yeah.

And this a whole bunch of work, uh, workload use cases for that, what I call in transaction ai. This has being able to do fraud scoring in the speed of the transaction flow rather than sampling 10% and doing it off off platform. You can do a hundred percent by doing it on platform.

So, So that's the hardware and then the software, Well the software case, then you've got the big vendors like A-B-M-C-A, Broadcom, a Rocket, an IBM, and a whole raft of the other vendors that are there. They're able to leverage and build off that hardware, but also they're accelerating with ai. So we're seeing code explanation, we're seeing AG gentech come to, um, operations.

We've heard today from Broadcom as part of the, what they're doing with technology, like open telemetry. So I think if you're to look at where this platform is today, we've just celebrated the 60th anniversary. We've just celebrated the 10th anniversary of Linux on the platform.

Wait, uh, that's what I was gonna get to. You mentioned Linux on mainframe. We've heard a lot today about IFL as well.

Talk to us a little bit about running things that aren't Set Os set os on. So We've just had the 10th anniversary of Linux one. I was one or two product managers who brought that to market.

Linux on the mainframe can go back to 1999, but back in 2015 we supercharged that. It's been one of the fastest growing bits of IBM Linux one, um, over the, over the last 10 years. Just the growth rate.

I can't share the numbers, but I I, I know them from my time and I I've kept track. We're talking double digit growth year on year. So this is the ability to run Linux on the S3 90 X chip architecture, which runs on tellem two, and be able to leverage the huge performance over five gigahertz performance, single chip, be able to leverage the massive memory footprint, but also tap into the availability.

This is the only server that's earthquake tested before it ships. So being able to tap into a box that is designed for seven nines, eight nines of availability, IBM quotes eight nines of availability in the specs. And they're a very conservative organization.

So to get a legal approval for a claim, they absolutely know they can stand behind it. When you look at that and contrast it with the cloud, talking about three nines availability. So just being able to contextualize, yes, this is an edge case.

Not everybody should buy LinuxONE and run games on it at home. We're not talking about a mass market server. The way I would describe it is if you want to do a journey, you've got options.

You can go in your car, you can go on a plane, you can go on a train, you can go on a boat. Just because a car exists. Yeah.

It doesn't mean we don't need a freight train. And the mainframe is the freight train of Yeah, This is your tanker that's gonna carry a big workload and be able to do it reliably at speed. Mm-hmm.

Now maybe that's only 10% of your workloads. That's fine. It's not a mass market server, but it does service a need that other boxes can't do in a way that they can't do it either.

And mainframe was one of the first platforms that had, um, uh, well LPARs virtualization, uh, essentially containerization as well. And, and that's what we're seeing a lot as well is that people are running multiple things at once. Yeah.

So if you look at Z vm, I think it traces its roots back to the early seventies. Um, so the virtualization, the ability to multiple layers. So you can create a logical partition that is carving a box up into air gapped partitions.

Then you can run virtualization on top of that. Can either run said VM from IBM, which is the license product. Or you can be looking at something like KVM or you can be looking at something like Red Hat OpenShift with virtualization options.

Then you can be putting containerized workloads on top of that. From a Linux point of view, there's even now in z os containers with Z os container extension. So you can be running Red hat containers within the Z os operating system, which just blows my mind.

If you go back seven or eight years, if you'd have said that that set of words, nobody would've believed you. So just look at open source. I got some this week that Zoe, one of the open source projects that sits under the Open Mainframe project stewardship is the 31st most vibrant software code base and the 53rd most commits of all projects.

That's 1500 projects that the Linux Foundation looks after. So there is a vibrant active community, and we're talking about stuff like Kubernetes and Open Telemetry and the Linux. That was the next thing for me to get to is that, I mean, everybody's talking about Ansible, everybody's talking about DevOps, everybody's talking about main, you know, they're not just talking about mainframe modernization as a as meaning, oh, let's get rid of the mainframe.

In fact, they're, what they're talking about is integrating the mainframe much better into the modern app Dev, CICD world. Yeah. So I think I take a pragmatic view.

I work with the types of A-W-A-W-S and other people. I think what we should start with is what is this application doing? What are its non-functional requirements for performance, availability, security, scalability, whatever they are, and then ask what is the right platform for that application to run a, if you answer that que set of questions sensibly and honestly, you'll come to a conclusion that some of those applications shouldn't be on a mainframe that are on a mainframe.

And you'll also come to the conclusion the other way round to say there's some applications that should be on the mainframe and maybe aren't. So I think what we, you should take the emotion out of it, look at it as an enterprise architect and go, what is this application trying to do? What are its non-functional requirements?

Give it a score out of a hundred out of 10 or whatever you want to grade it and then be able to go, where should I run this? A lot of applications should run on the cloud. A lot of applications should run OnPrem, some applications should run on the mainframe.

I think you've gotta be that type of balance and measured. And it Does seem, you know, like you said before, you know, maybe it's only 10% of your workloads, but a a lot of times for some businesses in some sectors, it's 90% of the value. I mean, and, and, and again, just to, I guess to sort of wrap up this sort of understanding what the mainframe is about.

Uh, what are those industries that are still mainframe reliant, and I don't mean still in sort of a negative sense, but that are, are in a positive sense that are basically we're gonna continue to run on the mainframe Something that's transaction heavy that needs state, the, the mainframe is typically in a place where state full applications. So we talk a lot about stateless, we talk about a lot about serverless, the sort of highly dynamic types of applications that don't care about their hardware. Fantastic.

Running serverless on AWS with some of their services. Fantastic way for you to go for certain applications. There's some applications where infrastructure does matter.

Those are the ones you should be thinking about. And what I'm hearing about is like latency sensitive applications, stateful applications, like you said, A database, a transactional system, those types of things. Well, And specifically things like airline reservations.

They all, well, all the successful ones that don't leave you stranded, run on the mainframe, you know, credit card processing, all the world's money runs through mainframe. Yeah, yeah. So there's, as I say, we're talking about a special use box that does certain things in a different way.

There's some trade offs there. But if you need that robustness and that security and that availability, there's certain things the mainframe does in different ways that other platforms don't do. So I don't look at it as one platform is better or worse than another.

You know, I've got relationships and work with the AWS guys. I've got relationships with HHPE and Lenovo and Dell and those guys. And I also work with IBM and a bunch of my, for me, I take a pragmatic view of what are we trying to achieve.

It's like, did, how did you get to Cleveland? Mm-hmm. You could have asked that question.

You drove this morning. I had to drive, then catch a plane, then catch an Uber. I took three different forms of transport.

I owned my car. I got on a plane with a multi-tenant service mm-hmm. But a different type.

And then I used a single tenant rented model for an Uber. All were valid for the journey that I was taking. Mainframe fits in that context of what are you trying to do as a workload and what are the non-functional requirements.

Yeah. And, and as a, as we've talked about, um, in many cases, these applications are made of traditional, uh, zeto s infrastructure components like, you know, DB two and co ball and all this kind of stuff, but also modern stuff like Linux and you know, Apache and you know, The IBM's done a great job of bringing pie torch and a whole bunch of AI languages and frameworks to the platform. You know, we've got, you can run an Elk stack on a Linux one, but I mean, all of this.

So I think painting the mainframe as this 50-year-old cobalt platform is just a bit disingenuous for me. Well, Thank you so much for kind of giving us that sort of understanding of that, um, and also of your special place in this community as kind of, you know, standing in both worlds and able to talk to the mainframe people in mainframe language and talk to the open systems and, and modern application people and kind of help, help to join them. And it's great to catch up with you at Cher.

Fantastic. Thanks for the opportunity as always, Steve.