A Conversation with Randy Kerns: A Look Back at 50 years in the Business – Infrastructure Matters, Episode 25

Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this webcast. The author does not hold any equity positions with any company mentioned in this webcast.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.

Transcript:

Camberely Bates: Welcome, everyone to 2024 and Infrastructure Matters. This is our first podcast of the year, and I am thrilled to have Randy Kerns joining us from Bonaire and his dive trip. I think everybody knows that he’s retired now, quote, unquote, “retired,” which that will never happen, but here he is, and we’ve asked Randy to spend some time with us and reflect on the last 50 years of his career and the industry. Randy, welcome.

Randy Kerns: Thank you. This is actually kind of fun.

Camberely Bates: I think we have you in between dives, right?

Randy Kerns: Correct. I’ve been down already, got up about forty-five minutes ago, and I’m going again in about an hour and a half.

Camberely Bates: Did you see anything really cool today?

Randy Kerns: I saw a bunch of Spanish lobsters, which was unusual, and I saw a number of these worms. They look like Millipedes, but they’re underwater and I saw them down around forty-five feet, so it’s interesting.

Camberely Bates: Cool, cool. I’ve known Randy for… Gosh.

Randy Kerns: Too long.

Camberely Bates: Yeah.

Randy Kerns: Too long.

Camberely Bates: I know we go back to 2022. Sorry, 2002. There we go. I got it spit out now. You’ve been in this industry now for 50 years.

Randy Kerns: A little more than that, actually.

Camberely Bates: Is it, really? Okay. Well we won’t go… What we wanted to do today was to talk about what the 50 years in data looks like, how you’ve continued to stay so relevant, and probably one of the biggest questions is why in the heck did you stay in data storage for 50 something years? Because at one point in time I said it was the most boring thing in the world.

Randy Kerns: Yeah, yeah. Just didn’t take much to prove you wrong. Seriously though, it’s been one of these areas that’s continually changing. Lots of new things happening all the time, and it’s always something to learn. We were talking about when I got into this, the first real way I got into storage, was actually in 1971 or ’72 in that I was working in a data center in St. Louis, and they were going to be developing a new application. They wanted to know how to optimize their performance when they accessed this information from their disk drives. These were CalComp disks in those days. So, I started looking into them and I started learning about count key data and gap one, gap two, gap three, and the time that you had to transition the gaps before you could turn the read heads on, and trying to figure out how to optimize accesses for this information from the disk drives we had. Yeah.

Camberely Bates: What the heck is a gap?

Randy Kerns: In account key data architecture, they actually have different types of what are called gaps, where there’s no data recorded, and you have time to control the electronics, and before you turn the reader right heads on again, before you go over an area where you can read or write, and that gap is a transition. A new, when you do a servo mechanism to move to the next one, you wait till you get oriented on a type of gap, to know where this start. Because remember these disk in the disk storage world are concentric rings, and so you have to deal with angular velocity and all these things. So, you get where you’re oriented, and then you have a certain amount of time before you can turn the right heads on. Then you can get the count field, and then you can have a key field, which is an index if you will, and then data, and you have a certain amount of time for each. If you don’t do things correctly, you have to go complete the revolution before you get to that same point again to read the data. So, it’s very critical.

Camberely Bates: It’s a whole lot different than using solid state and flash, isn’t it?

Randy Kerns: Well, if you remember, I went into Future Systems in IBM, and we were working on the… Well, it’s going to replace the System/370, and we saw all these problems and we said, “Okay, well let’s change to a fixed block architecture.” So, we could just deal with an extent that says, “Here’s a start and there’s so many blocks there and you’re all fixed size.” So, it simplified everything and certainly made performance more predictable. But eventually we didn’t do that for the next generation 370. We’d actually done that on, we created a fixed block architecture 3380, and then we had to go back. I was involved in that group, we went back and put count key data onto 3380, and the stuff we’d been doing for that new system actually became the System/38 IO system.

Camberely Bates: Which is still one of my favorite systems.

Randy Kerns: Yeah, it’s the AS/400. It evolved, and it’s still fundamentally the same. The only difference is we’d had a unique 48-bit key, and it got expanded to 56 before it finally was released to the field. It’s a very unique architecture we worked on back in the early seventies.

Camberely Bates: That’s the seventies, eighties. You also went off to start working on this incredible system that was great.

Randy Kerns: On what?

Camberely Bates: Working on this incredible… Boy, those trade winds are wonderful. Just nice and-

Randy Kerns: Yeah, sorry.

Camberely Bates: Also worked on this one incredible system that was way before its time, called Iceberg at StorageTek, and you were the lead executive of that, if I recall right.

Randy Kerns: Well, actually, I can’t take credit for that. I did get responsibility for it at one point in my career, but I had actually been brought in from… That was my second stint at StorageTek, and Jesse and Dave Weiss hired me to come back. It wasn’t Jesse this time, it was Dave Weiss, convinced me to come back and go to work there. They bought another company, and we were doing another count-key- data architecture storage system. But the concept with that one was to be able to move into open system storage, so move away from count key data to fixed block.

Iceberg was being developed several years ahead, and it was very embedded that it was only going to be count-key-data architecture, but it was incredibly novel, way advanced, and the guys that did it, incredibly bright people. So, they’d come up with this, but it was going to be limited. So, we did this deal with IBM at the time to resell, both the Iceberg product and the Kodiak product, and Iceberg did well with customers. At one point in time they had us change the back end to a serial architecture from the parallel architecture it had. So, I got put in, responsible for the whole program for doing that. So, that was my part of it. As soon as that was finished, that’s when it was time to say, “Okay, no new programs, time to move on to something else.”

Camberely Bates: You then went on to open systems, and some great stuff over there. Tell me how open systems was so different, maybe different from AS400, what you were doing there, what you were doing with the mainframe systems, et cetera. What was that going there? What was the big shift?

Randy Kerns: Well, the big shift is those were open, from the standpoint that lots of people could contribute. Lots of proprietary software was added, proprietary drivers. There was a degree of openness when you could have one guy somewhat controlling the OS and then a lot of people wanting changes and trying out changes. So, it was a continually evolving area. From a storage standpoint, there’s only so much you can do physically with the storage once you move to a fixed block architecture. But there was a lot of interesting things you can do around caching, around buffering data, creating some type of environment where you can ensure consistency with a network-type access. That’s where SANS came in ’99, and that allowed us to get more utilization out of the storage more than anything else and do a high degree of parallelism that we didn’t have before.

So, all those things kind of came about nearly the same time. Open systems kept expanding to the point today we’ve got lots of open source in all kinds of different areas, and that’s certainly been beneficial, but it’s also been somewhat problematic in that it means all that code has its potential issues exposed to the world to be exploited by people for nefarious reasons. So, it makes it a lot more complicated.

Camberely Bates: Does that mean that it’s more exposed than what is being done on the mainframe nowadays?

Randy Kerns: Oh, absolutely. Absolutely. Yeah. People can go in and find weaknesses, the Log4j thing, for example, and so there’s more opportunity. Now, normally in the storage world, the things that happen in the storage system aren’t really ever penetrated, if you will. It’s more from the application layer above and things are done to it, like encrypt data, delete data, and that’s where you end up doing a lot of your protection mechanism in the access to the storage. You certainly see it a great deal in the data protection software. Things trying to make sure that, since it controls and knows where all the protected copies are, protecting those copies is utmost important.

So, in case there’s been something else happen, you can get your data back. But you see the applications and the other things being focused on from a protection standpoint. Not the devices itself, other than the fact that we’re exploiting things to be able to do what’s called an air gap. What that means is that you can write some data and then you don’t provide a programmatic way to go after, that an application can get to.

Camberely Bates: So, mainframe, future system AS/400, open systems, cloud. We’re putting stuff in the public cloud now. Have the decision criteria changed at all from the very beginning?

Randy Kerns: Well, if we narrow it to storage and more importantly, information, information, storage and protection, the problems are actually still the same. They go back to the sixties, seventies, et cetera. I want to access it quickly. I want integrity. I want to be able to protect it. I want to ensure that I don’t have inadvertent access, all these other things, and then I want to do it for the least possible cost while still meeting these other criteria. Those were all present when I first started. Those haven’t changed. The mechanisms have changed. Certainly some of the requirements have gotten more intense, if you will, and more varied. But the fundamental need for what we would call the customer, the IT organizations, they’re still the same.

Camberely Bates: Why do you say more intense? Why is it more intense? What’s caused… Yeah, I’ll leave it there.

Randy Kerns: There’s two things. One, there’s always a screaming about wanting more performance. Wants to ramp up, and-

Camberely Bates: What I thought solid state was supposed to just make all that happen.

Randy Kerns: Oh, it made a world of difference, but what it did, is it moved where the bottleneck was or where the limitations were from the device access to the network, and then from the network to the data mover functions done in the servers. So, the solid state data was fabulous in that it made the performance tuning simpler, limited the amount of effort required to get data quickly, but it exposed the other bottlenecks which have been addressed. We’ve seen that with NVME as a protocol now, to get rid of the command response interlock we had before.

We’ve seen it in the server systems now doing these offload controllers, if you will, these DPUs, data processing units, and other things to try and get the processing function away from just handling the movement of data, because that was the bottleneck. So, we continue to see these things changed, and so solid state did make a world of difference from the performance standpoint. There’s never enough there. But to answer your question, the networking of everything, if you will, blame it on the internet, that’s a easy one.

Camberely Bates: Well, blame it on Cisco or something like that.

Randy Kerns: Nah, those poor guys. The networking of everything has created opportunity for these attacks. Now all these attacks are saying, “Okay, how do we prevent this?” Really, your last line of defense is to restore data that you’ve made a protected copy of. So, that’s brought the focus into the storage area, the management of that, the security of that. And that’s where that intensity has come in, because that really is the last line of defense. You can put all the other avoidance and detection things in, but there are people trying to always figure out ways around them, and it’s tough to keep ahead of them sometimes, but you ultimately resolve to getting that protected copy, whether it’s S3 object lock on object storage, or whether it’s even a tape that’s been removed.

I gave you the story one time that, when I was in a startup company, we brought in a guy to run the IT operations that I got to know real well. He’d been flying SBDs to the Navy right after World War II. He was a really funny guy, but he was one of these guys that, at night, every once in a while I’d seen him taking several tapes out, and these were the big, round tapes, to his car. I said, “What are you doing?” He says, “Oh, I keep a copy of backup tapes at home.” He said, “I’ve got a refrigerator in the basement that’s not plugged in.” He says, “We ever need to get data back, I got it.”

Camberely Bates: Okay. So you have to tell that one story though about the repairing of the tape down in Colorado.

Randy Kerns: That was a disk.

Camberely Bates: Oh, that was a disk. Okay. Got to tell the story for folks. This is a fabulous, fun story.

Randy Kerns: Okay. Well, I was working for IBM and one of the guys that was working in the Denver field office fraternity brother of mine, so I knew him real well. He knew I’d been working on the 3350 attachment to 3850. We were making a switch over, it was called Madrid Staging, was the project. And so I obviously learned a great deal about 3350s and the embedded code for them and all that. So, they were having a problem in the, let’s just say Cheyenne Mountain.

Camberely Bates: Cheyenne Mountain is… Exactly what is Cheyenne Mountain?

Randy Kerns: Everybody knows. Anyway, he was saying they couldn’t figure out what it was, and these were the symptoms. And I said, “Oh, I know what that is,” because I’d seen it before for. I told him, he says, “Okay, well, I’ll tell the FEEs, and apparently they wouldn’t let the FEEs in to repair it, but they disconnected it and rolled it out. They put the fix in, and rolled it back in. Just-

Camberely Bates: We do have to explain Cheyenne Mountain because there’s going to be some Europeans that are watching this.

Randy Kerns: It was a secure facility, let’s just put it that way.

Camberely Bates: And literally in a mountain. Literally in a mountain. So yeah, down in Colorado-

Randy Kerns: Yes, that was kind of interesting, and it just so happened we were talking about this, because he I’d worked on these things.

Camberely Bates: That is great. That is a great story. We kind of touched on this, but maybe a little bit more succinct in it. What do you see in your 50 years here, what were the biggest shifts?

Randy Kerns: Well, you brought up solid state. That’s been one of the biggest ones, and I just keep telling this to people, you can’t underestimate the value of moving away from electromechanical devices. When I started, we had key punches and card readers, and one of the things you had to do when you first start, is you have to learn to fix your own key punch, because you’re going to be working at 3:00 A.M. and a repairman ain’t going to come. So you learned how to fix all these, and these were all electromechanical devices. If you get away from the electromechanical, you can be so much more productive. So, that’s what’s been happening with solid state. That’s really a really big deal.

The other ones are maybe outside of the storage area, but they have to do with development. Development’s really changed. When I first started, you had maybe 64K in memory, K, and so you wrote a lot of elaborate software to be able to page in pieces of code to execute, and then you had to do self-modifying code in some cases. Now we don’t even think about the size of memory we have unless you’re the database guy. Now all of a sudden we’ve got CXL coming along that’s going to give us massive, massive amounts of memory, effectively. So, the idea of managing and doing IOs, the things I was talking about earlier about count key data, index sequential operations and all, that all goes away. So, now you treat all of… The fact that you have data, it’s all amorphous, and then the system handles that, putting it into memory and handling the IO to simplify everything.

So, the big changes are this change to infinite memory, if you will, the CXL that’s coming, the change, getting rid of electromechanical devices, that’s a big deal. Then the development tools, dramatically different. When I first started, you had this triumvirate. The big thing in my career was to be able to say, you designed your own processor, you wrote your own assembly language for it, and you wrote your own high-level compiler for it. So, I had done that. Nowadays if you talk to anybody about doing that, they look at you funny. Just those things aren’t necessary anymore.

There’s so many great tools that are available, and the programming languages are so much more powerful. Sometimes they’re a little too wordy and verbose, and some people misuse them. Anytime somebody says YAML to me, that’s an example of somebody misusing them. But the point is things have gotten much simpler. One of the things we’re starting to see now is this transition away from having to write your own scripts, write your own APIs, et cetera, to these low-code or no-code tool sets. And that’s going to again, expand the horizon to where you have less specializations, more… You’re going to get more out of the people. It’s going to be very interesting.

Camberely Bates: Wow. Okay. So, let me shift in the last section here. Given all the current industry shifts, you’ve been working with the data architects or the storage architects. Now we’re kind of talking information architects, folks changing their titles a bit. Your advice to the current people responsible for information and data infrastructure. How do they stay, A, relevant? Because a lot of people say, “Oh, you don’t need to manage this stuff anymore, et cetera.” What’s your advice to those people?

Randy Kerns: Well, there’s a lot of people say that, but you’re more valuable if you really, truly have an in-depth understanding. You’re much more valuable. You can explain when there’s questions asked. More importantly, you can figure things out when something goes wrong, and invariably things go wrong. So, if you are at that superficial level, then your value is not there, compared to somebody that has that more in-depth knowledge.

Now, the way you get your in-depth knowledge is difficult. We do a class. I think that helps a lot of the current IT people to broaden their horizons, understand how things work and what’s different and why there’s different value from different solutions, et cetera. But working in it is one of the better ways. One of the things I always advocate, is for somebody to really understand something, they need to understand to a level where they can explain. So, if they can take any particular area and go into it, do the homework, go to a class, read about it, and then be able to maybe teach a one-day class on it to somebody, that’s a demonstration that they really understand it. In our industry, we call people that understand something subject matter experts, and that’ll get watered down eventually. But right now it’s normally a badge saying, “This person has done their homework, they’ve lived it, and they really understand it.” So for right now, it’s a good area. It used to be that there were other people had specializations or knowledge, and we gave them different titles and things, but those tend to get diffused over time.

Camberely Bates: Yeah, and I will segue and do a bit of a big commercial on the class that we’re doing, which is April 16th and 17th. We’re going to be sending out information on it next week for registration. It’s the last one we’re going to do.

Randy Kerns: In the US.

Camberely Bates: In the US. Well, you’re right, we’ll be over in Europe for our buddies over there. You get two days of this, two days of this type of knowledge going and being delivered in that class. Some of the cool stuff we’re going to be talking about this year, is all about what we’re calling the AI data platform, which is like, if you’re involved or you want to get involved with generative AI, guys, this is the next generation of where this data is going and what’s happening. By the end of this next year, we’re going to see this kind of like a rocket- IT organizations. So, anything you want to say about that class?

Randy Kerns: Yeah, we’re going to talk a lot about that, but it’s not that you want to get involved. I think a lot of the IT guys are going to have it thrust upon them. So, they really do need to understand the framework, what’s going on, what their role is going to be, what the issues are, and where there’s real value. Because it can be quite overwhelming initially. A lot of the people that they may be working with don’t come from an IT background. So, there’s some terminology, and I can’t help in teaching patience, but it’s something you do need.

Camberely Bates: Oh, by the way, Randy is coming back from Bonaire for this. He’s not staying there for the entire-

Randy Kerns: No, it’s just this week. You run out of money after a while here.

Camberely Bates: So, we will be in Boulder, Colorado for that in April. So we would love to have you guys all here and just kind of take over the hotel that we’re doing at the Boulderado. So, we will welcome you. Randy, thank you so very much. We’re going to do a couple more of these, I think, over the next month and see if we can pry a few more things out of his brain. The next one that we’ll be doing, is going to be covering some of the mainframe stuff, which is… We’ll send that over to all of our mainframe buddies that we love and are dear.

Randy Kerns: Okay.

Camberely Bates: Okay, thank you very much guys. Don’t forget to click and follow. We’re going to put a link in here so you guys can come over into our class. We’re going to love to have you guys. Randy, thank you so very much. Go dive.

Randy Kerns: Alrighty.

Camberely Bates: Go have fun.

Randy Kerns: All right. Bye, guys.

Camberely Bates: Bye.

Other Insights from The Futurum Group:

Join the Information and Data Management; Strategies and Solutions class, April 16 – 17, 2023 in Boulder Colorado

Talking Cloud Cost Economics with Senior Strategist Randy Kerns – Infrastructure Matters, Episode 16

What Is Moving the Acquisition of Data Infrastructure?