#395 The Role of Data in AI Innovation: Erik Salo’s Take on HPC and Storage Systems

[00:00:00]

Mehmet: Hello and welcome back to a new episode of the CTO Show with Mehmet. Today I'm very pleased to join with me, Erik Salo, who's the VP for Marketing and Product Management at Vdura. Erik, thank you very much for making the time today. I know like, for someone like [00:01:00] yourself, how busy it could get, but you took the time.

Mehmet: So what I would love to do is, you know, to keep it to you, to introduce yourself, tell us a bit about your journey, and then we can start the discussion from there.

Erik: First, it's a pleasure to be on the show. So I really appreciate the invite. So let's see. So, um, you know, I started out, I wanted to be a mechanic, right?

Erik: I wanted to, I love machines. I wanted to work on cars and things like that. Parents told me, they said, you know, you got to go to college. And, um, so I ended up, uh, I ended up in the computer business almost immediately. And man has it been interesting and fun. Uh, you know, my first job was, um, as an engineer at a power plant.

Erik: And that was like a summer job and then my first real job was actually As an engineer at a company that made quarter inch data storage cartridges So they were called carlisle memory products and they made these, you know, little little little data tape projects I think they were maybe 60 megabytes or something like that.

Erik: And you know, it's crazy. This is you know, 30 something years ago And when I joined the company, they [00:02:00] said, Hey, welcome to this company. You know, it's a great place to start your career. You're really going to grow here, but you know, tape not really going to last more than three or four more years. And you're going to have to probably do something else, you know, for your career, you know, it's amazing.

Erik: 30 something years later, people are still using tape, but that was my, my first job was, was in the data storage industry. So super fun, super interesting. Uh, I went to AMD after that and man, what a great run I had at AMD. Uh, more than 10 years and, you know, some of the good times we had the first 64 bit processor while I was there, I started and ran the embedded division.

Erik: It ended up, um, being the, the strategist for the desktop group. Uh, I was there for the acquisition of ATI, participated in that, it was really a good move. You think about it today with AI, you know, if, um, if AMD didn't have ATI, it would be a whole different, whole different company. I was also there when they split from having their own fabs, you know, originally AMD like Intel had their own fabs and they said, Hey, we can't really afford this.

Erik: And so they split their global boundaries and, you know, both of those moves, you know, this many, [00:03:00] probably 20 years later seem kind of prophetic. I did a startup after AMD, which was super fun. Um, uh, cloud software and technology for electric vehicles. I loved it. It was three years. Didn't go, but man, what a great group of people and a lot of fun.

Erik: And then I went to Seagate and I had a great run at Seagate. Uh, I ended up, um, running all product and markets for the OEM business and the, um, the systems business and the distribution channel, so a couple of billion dollar portfolio. And then about a year ago, I, uh, I moved from a big company and moved to, um, the Dura, which is just an amazing, uh, change.

Erik: And, you know, we make, we make software storage software for, uh, for HPC and AI. And it has just been a blast and the opportunities we have to do the right thing and to make great products are really exciting. So there's the, there's a two minute, uh, summary of my whole career.

Mehmet: What I write, I would say, uh, Erik, like really it's a very, you know, great experience and I'm, [00:04:00] I'm sure that the audience will benefit a lot from, you know, every single what you're going to say today.

Mehmet: So it's interesting that, you know, like it looks like you, because when we talk about machine learning and AI, we still talk about data. So, um, Just, you know, I'm always curious to, to, to speak, you know, especially from products for someone who does the product and the marketing also as well. So when it comes to machine learning and AI, I think there's a lot in the back end, as we call it, that doesn't come to the picture, which is, you know, the importance of, of having a proper data platform and, you know, like having your data set in places.

Mehmet: So if you want to, to, to walk us through from. You know, I would say, because I I have a, you know, a little bit kind of, uh, uh, special treat to to everyone who works in product. So when it comes to data platforms, like if someone want to explain [00:05:00] there was there will be no one better than someone like yourself.

Mehmet: I would say Erik to explain to us, you know, the importance of data platform when it comes to machine learning and the eye. So why it's it's important to have this base if you want.

Erik: It's such a great question. And, you know, Data is where the gravity is, right? So, you know, I spent, you know, fundamentally half my career making CPUs and designing CPUs and half my career making storage.

Erik: And the gravity is with the data. And that's the thing that's, that's the critical insight. You know, if you have a compute instance, you can spin it up in, you know, in AmErika, and then you can turn it off and spin it up in Europe somewhere and, and, you know, you can, it's, it's somewhat ephemeral. It's not trivial.

Erik: Right. There's a lot going on, uh, especially with. The A. I. World, you know, the computing, um, strategy has changed significantly. But the fact is that it's all about what you're computing, right? And you're computing data, and that's the whole thing. And so the and the data is hard to move, right? It's like, you know, if you have a [00:06:00] petabyte of data, oh, my gosh, that is a that is an amount of I.

Erik: P. That has that has its own gravity, and, you know, it's not easily just move back and forth between between places. And so the Strategy for creating data, keeping data and accessing data is really the key to all these modern innovations, right? If you look at it just from 30, 000 feet, it's actually a revolution in what we're doing with the data, not with the computer.

Erik: It's very interesting. Of course, they're directly related, but the data is the thing with the value, ultimately, and it's the thing with the gravity.

Mehmet: Absolutely. Now, when, when we talk about data, uh, Erik also as well, and we know how important is the data for, uh, you know, training the models and, you know, like making sure that, you know, these LLMs, you know, have this. So if you want to walk us through also like this specific use case when it comes to, um, You know, [00:07:00] uh, training, you know, for, for, uh, of the data for the LLMs and these use cases that everyone start to talk about.

Mehmet: So like not technically, of course, but you know, on, on a maybe high level and maybe you can touch base also on, on how, you know, you can enable this when, uh, you talk to customer in, in your, uh, role today at the Vdura.

Erik: Sure. So, you know, one of the things to realize is that all of the modern, let's are happening so fast.

Erik: In this world of artificial intelligence and machine learning, you know, it's like it seems like you're You're you know, your paradigm is obsolete in a couple of months If you don't if you don't pay attention, but fundamentally all the current models here in september 2024 are Based on this strategy from this google paper attention is all you need and what they do is they get this big bunch of data And it's all about the data and they say all right Well, let me you know, let me let me give this thing a prompt and let me let me guess an answer Um And how close is the answer to what you know my data sets right and just does that over [00:08:00] and over and over again.

Erik: So let's think about what that means. The simplicity of it is striking by the way. It's just striking how simple this is now. It's not easy, right? But what's happening is quite simple. It's just happening at a scale that I think any human has trouble comprehending. But. If you think about what's important.

Erik: So first of all, you've got to have a big data set, right? I think that, you know, when we, when we look back on this industry in 10 or 15 years, and it's a little more mature, the winners are going to be the ones. Not with the big compute, but with the good deep data sets, I think that's really the thing that's going to make a difference because ultimately, you know, these, all these machines are just basically guessing what the next, you know, what the next bit of data is going to be.

Erik: So if you don't have any data, you can't do that. So what is having the data? And two is, Keeping the data. And that's one of the things that we really focused on at Vadura is that, you know, the storage that is close to the compute is always very fast, right? [00:09:00] And, you know, even, you know, for many years, you know, it's funny, we talk about AI being a new thing, but I've been working on machine learning stuff for, you know, pushing 30 years, right?

Erik: A long time. And you know, we called it other things and I think, you know, it was more, it was more boring, but ultimately you know, this idea that, that computers look at data in a different way. In a thoughtful way and find patterns in the data that are useful to people is ultimately what's happening here.

Erik: And I think it's happening at a level of effectiveness. That's astonished the world, but ultimately it's been going on for a long time. This is, you know, we're not. This is not the beginning, but you know, it's a we're decades into this into this journey. So the storage is very fast. That that talks to the computers and in general in the beginning, right?

Erik: When people created parallel file systems, what I call parallel, meaning like many storage elements, usually hard disks, many storage elements work at the same time, and when you add up all of their transfer rates, you get really fast data, right? So that's called parallel file system and the initial, the early parallel file systems like luster, [00:10:00] GPFS, they were really fast, but we're really Uh, kind of glitchy, right?

Erik: We always call it like I work a lot on on luster earlier in my career and we called it like an F1 car. It's like, you know, it goes really fast, but it takes a team of engineers in white coats to start the thing and it crashes quite a lot, right? And that's, you know, that's how it's been. That's how high performance storage has been for a long time.

Erik: Um, what? The original vision for the company that we have kept is like, make storage fast, but also make it easy, right? Make it reliable, make it easy to use, just like enterprise storage, like a NAS or something. And so, you know, as the, as these LLM models, and in general, what we're doing with machine learning, as that matures, It's going to become more and more important that your data is safe, right?

Erik: And that's our number one. That's our number one goal of the company is keep customers data safe. Wake up every morning. Think about how do you not lose your customer's data? Because ultimately I think that's the valuable quantity. That's the thing that's going to distinguish [00:11:00] who wins this race with, with AI is who's got the best data, who can keep it safe and who can use it effectively.

Mehmet: Absolutely, Erik. And, you know, out of curiosity, because you mentioned, uh, you know, about safeguarding the data and, you know, it's on top of mind. So what are like some of the, I would say, shortcomings or challenges you see in the customer space today that they are maybe not, I would say, putting a lot of effort in doing or maybe there's some, uh, I would say the lack of especially when it comes to, you know, to I know this for a fact from my personal experience, but I want to see also your point of view.

Mehmet: So because, you know, the customer thinks usually that, you know, this data is replicated and, you know, if they are using, let's say, data lakes or they are using something in the cloud or something like this, so they think it's [00:12:00] not part of what they should be doing. So, If you can enlighten us on that part, because I think this is very important, because if you lose the data, you lose your ability to do anything right again.

Mehmet: This is not only for for the A. I. Use cases, but I think with A. I. It becomes more critical and crucial to safeguard this data. So what are you seeing in that space, Erik?

Erik: Well, you know, one of the amazing things is that the technology that stores most of the data used by A. I. Today. It's based on a very old strategy, right?

Erik: So you think about, so it's like, let's go way back. I don't know how far back, but let's go back before clouds, right? Clouds are old.

Mehmet: Sure.

Erik: What, what people did was they did what's something called highly available architectures, right? Dual, dual active. And what they do is they have two computers, essentially.

Erik: A and B and each computer could talk to all the stuff that the other one could talk to. Right now, normally they each talk to say half the, the storage elements of your elements, but if one of them broke, right, if one of them failed, then you could, [00:13:00] then you, then you could fail over to the other one. Right.

Erik: And that's called scale up. That's a scale up traditional architecture. When I was at Seagate, you know, we sold Santa Rays, sold probably more Santa Rays than anybody in the world. Man, I got, I've been selling those products for decades and decades and decades. And they're pretty good until they start to get big.

Erik: Right when they're small, right when they're, when they're in half a rack or something like that, which was a lot of storage 20 years ago, all of a sudden that strategy starts to break down. And as those systems get larger, they actually get less. Reliable, right? They get less durable, meaning they're more likely to lose data and they get less available, meaning they're, you know, they're more likely to go offline and, you know, give you no access.

Erik: And so as in the early days, you know, when the, when the first CSP started thinking about this, they're like, Oh my gosh, there's no way at the scale of data we're thinking about, there's no way to make. A single or dual scale up device reliable enough, you know, it's like if I've got one of them, that's pretty good.

Erik: But if I've got 50 of them or [00:14:00] 5000 or 50, 000, there's no way right? My numbers just go down, down, down. And so they came out with this other strategy to scale out strategy or shared nothing strategy where a cloud basically has all these different elements. Let's say have you heard of erasure coding? It's a subset array, but it's fundamentally the, um, or the other way around.

Erik: But it's the, it's the way that we protect data in these large systems. What you do is you take a chunk of data, say a picture of your cat, and you say, okay, I'm going to, um, I'm going to break that picture of a cat up into 10 chunks. Then I'm going to have two other chunks, and if I lose any two chunks, any, any of them, I can get the whole picture of it, the cat back, right?

Erik: So that's erasure coding. And it's turned, turns out it's statistically a very, very, you know, depending on how much you want to erasure code, you can come up with, Durabilities that are astonishing, you know, 99. 9, I don't think. And that's what you think up in clouds. Unfortunately, a lot of the high performance storage has not done that.

Erik: They've stayed in [00:15:00] this dual, dual active, active world. And one of the things that actually attracted me to come to the Dura was that our strategy is we have a scale out strategy or we have a shared nothing strategy. So the bigger the system gets. The more reliable the system gets, the bigger the system gets, the more valuable data you have, the less likely you are to lose it, right?

Erik: And the more likely you are to be able to access it. And that's a fundamental architectural advantage that we have. And it's like, I think it's one of the biggest strengths of the company is that, you know, I think this is more and more important as this data becomes central to how we live.

Mehmet: Right. I gotta ask a question.

Mehmet: Of course, for me, I know probably the answer for that, uh, coming from a technical background. But Erik, I want you also to maybe explain a little bit. I think this is very important. And for someone who have been in this domain for a long time, like yourself. So we talk also when it comes, you know, to, especially to AI today and machine learning.

Mehmet: So we need the speed, right? And, you know, we talked [00:16:00] about high performance computing for, you know, Like decades, probably HPC's and now we are in an age where people start to talk about, you know, I want you to, to demystify this. So when we talk about HPC at the same time, we see now, you know, the rise of, uh, GPUs and TPUs and all this.

Mehmet: So how like all these intersect together in, in the world of machine learning and AI.

Erik: Man, that's a good question. I think if I could, if I really knew the answer to that question, I'd probably have a different role, but there has been a really interesting shift and it's happened, um, more slowly than you might think, but, you know, in the, in the beginning, you know, the HPC, they use these general purpose processors, which are incredible feats of engineering and they're really good at, um, they're really good at a wide range of tasks.

Erik: And what we found with some of these [00:17:00] really specialized. Um, tasks, right? Like I is that the highly parallel GPU type engines are just faster, right? And they're actually quite a lot faster, you know, quite a quite obvious. And I, um, you know, I've played around with the Google TPUs and there's a new one called Halo.

Erik: And, you know, these, these little things that are the size of your thumbnail, the size of a little, a little M2 chip, you know, they're doing, you know, 14 trillion operations per second for a couple of Watts kind of thing. And so, you know, the, you know, The difference is that to train these LLMs, they're very.

Erik: amenable or they're very, it's very efficient to use these GPU type processors. And that's, that's, that's the big switch is that we've, we've, we've kind of found a workload that is much better served by a specialized processor than a general processor. That's the big difference.

Mehmet: Yeah, I think this is fair enough.

Mehmet: I would say Erik, uh, well, at least for me, you know, this, this answer is, uh, [00:18:00] is, is, you know, very good. I would say now another thing which, uh, you know, I want you to touch base on, which is, you know, Some of the use cases is like beyond machine learning and AI. So we start to see a lot of talks about autonomous systems and systems that they can, you know, act on their own.

Mehmet: Right. So not scary at all. Right. Sorry. Not scary at all. Right. Absolutely. So So in that specific use cases, it's like autonomous systems. So are we talking about, like, something similar to, uh, predictive made in the space of predictive maintenance? Are we talking in manufacturing? Where exactly are we seeing the use cases is more in autonomous systems?

Mehmet: Well,

Erik: you know, let me let me go on a little bit of a tangent [00:19:00] before I answer your question. So almost 10 years ago, I realized That there was a fundamental thing changing, right? And it wasn't that people were making more data. People are making a tremendous amount more data than they ever have. You have to think about it's just like X to the cube.

Erik: That's not, it's not actually the big story, right? The big story is that machines have started making and spend some time now. Machines started making data for other machines, right? So it wasn't people making data for each other. It wasn't people making data for machines. It was machines making data for other machines.

Erik: And that's the key thing because, you know, computers operate so quickly that as soon as you get into this circle, you can all of a sudden, you know, get incredible growth in whatever it is you're trying to do, right? So that's the, That's the big thing, you know, one, one interesting insight is that, you know, it seems to me an unintuitive It's going to stay this way, but boy, it sure looks like it is When you train an LLM, you know You give it a bunch of data and you train it and then you basically just kind of train [00:20:00] itself after that and it so far it seems that the more Cycles people throw at these, these LLMs training themselves, the better they're getting, which is just astonishing, right?

Erik: Machines making data for other machines. So to actually get to your question, what, what industry is going to be most impacted? I think every part of our lives, every industry you can think of is somehow being transformed. By this data revolution. So you can look at it in manufacturing, you know, you think it wasn't, it's been, you know, before this big revolution, you know, some of the big, you know, big, big, uh, retailers were using machine learning, not now called AI to figure out how to stock shelves so that their robot pickers could go and, you know, more efficiently get, um, the stuff off the shelves.

Erik: You're seeing that, uh, but I'd say energy oil and gas. Uh, they are taking all this seismic data and other kind of surveyed data they have that's very expensive to create and the algorithms just keep getting better and [00:21:00] better and they, you know, they, they learn, you know, if there's this pattern of, of something in the data, then that means that there's, you know, some, some oil underneath there kind of thing that's getting better and better that is continuing to be revolutionary medicine, genomics.

Erik: I mean, I just go forever, right? Like there's no industry, That this is not touching and improving in a really substantive way. Right. And that's, what's exciting about it is I just can't think of any part of technology or our world where you're not seeing, you know, some, some big revolution and, you know, you're going to have self driving cars and, you know, the writing is, is great and video generation, you're seeing movies being made, I just made an avatar of myself.

Erik: And it's not perfect, but man, it's pretty good, right? So, you know, you're just seeing it every everywhere. It's this explosion.

Mehmet: Yeah. It's, it's fascinating. Absolutely. Erik. Um, one thing which I should have maybe asked you when, when we talked about, uh, you know, the storage and, you know, that part of, of, uh, uh, what, [00:22:00] what you do at Vdura, so I've started to see a lot of people saying, like, I can't remember what accelerated the cloud adoption, but people are saying that machine learning and AI are accelerating, you know, the data center business to, to, you know, start to grow again.

Mehmet: Are you seeing the same trends, Erik? Are you seeing like people are more tending to move their workloads or maybe the workloads that would be Used for their LLMs training or maybe I don't know like for their data sets to to be on on On their own data center. Are we seeing something like that?

Erik: You know, I am not seeing clarity there You know, it's funny I've been in this business long enough where i've kind of seen the everything to the cloud everything out of the cloud Okay, some stuff in the cloud, you know back and forth and it's not clear to me How this will go in the long run now Right now, nobody can buy GPUs except for the big clouds.

Erik: When you think about, um, [00:23:00] uh, you know, Meta, I think, just agreed to buy more GPUs, more money in GPUs than it costs to do the Manhattan Project that created the first nuclear weapon, or the Apollo Project that put a man on the moon, right? In real dollars, in today's dollars, right? The astonishing, so, the clouds are buying up a lot of the GPUs, so in many cases, the only way you can get them is to use the clouds, but, yeah.

Erik: It's just not clear to me. You know, there's, and there's a couple of factors, right? So one, again, the gravity is with the data, right? So if you're a company that has a bunch of data on your own, boy, it does not make sense to store it in the cloud financially, right? Although it might be a lot easier, et cetera.

Erik: And we do see some people doing that, but I actually don't, I think that you'll see almost forever this hybrid environment where when you've got a lot of data or a lot of compute that you want to have in your shop, then you're going to want the on prem. thing. And if you are either small or, you know, you got a, you're really geographically diverse, uh, disperse things like that, then you'll be in the cloud.

Erik: And I just think that forever we're in this hybrid environment. I don't [00:24:00] see, um, I don't see a clear path one way or the other. I'm sure the cloud people would disagree with me and the on prem people both disagree with me. They each think it's all coming to them, but I don't see it right. I see, I see a mixed environment probably forever.

Mehmet: Yeah, but the reason I ask this question because I think over, you know, and just for the sake of transparency, we are recording this on Monday, 9th September, so probably this will be out after two weeks from now. So during the weekend, I've seen, you know, uh, some, uh, posts from Michael Dell and, you know, he was saying like, yeah, so the business is back about the data center because it seems, you know, During the when people were shifting their workloads to the cloud, and now things started to flip around.

Mehmet: So people want their compute close to them. And this is something actually you just mentioned at the beginning, Erik also as well. So so that the challenge was always, you know, to to keep the data as close as possible to to, uh, to the to your compete to compute, right? So, um, Now, [00:25:00] if I want to ask you after, you know, you've, you've seen it all.

Mehmet: And to your point, also, you repeated this multiple times that AI machine learning are not the new, they, you know, the, the, the study started in the early fifties of last century, but of course, you know, generative AI, I think, you know, it, it started, it awakened the giant, I would say so, but from.

Mehmet: infrastructure perspective, if you can, you know, do some predictions, I would say, and you know, what do you think would be the next big thing from from the infrastructure perspective, of course, use cases you just mentioned, so there's no single, you know, Aspect of our life that would not be affected. And we're gonna see these advanced innovations.

Mehmet: But if I want to think about, you know, the infrastructure. So what are the things that you are exciting you? The more I would say,

Erik: Well, I'll tell you what, you know, 11 [00:26:00] big thing that's excited me and got me to move to the Madura. And one of our big bets is that there's gonna be a convergence between in the storage perspective, there's gonna be a convergence between the easy use reliability of the technology.

Erik: Enterprise world and the speed of the HPC world. So again, in a lot of ways, there's been two parallel storage worlds in the, in the kind of the high, high end, you know, computer storage, the fast stuff, the parallel stuff has been fast, but just really flaky, right? Flaky is the right word. It's just, you know, it goes down all the time.

Erik: It's hard to manage. You have to have a, uh, a group of people. Yeah. Storage admins to take care of it. There's a million knobs to turn sort of thing, right? And you know people have been willing to pay that price because they like their performance, right? So that's been one world and then the other world that's growing up is one And I spent actually most of my career here is the enterprise world where it those products have become just tremendously mature You know the kind of the nas and san world.

Erik: They're easy to use. They're incredibly [00:27:00] Reliable in protecting your data. And of course what that set of customers wants, they just want to turn the darn thing on and have it work. And they, you know, when they, when they need the data, they want the data and they don't want to have to have a full time admin to bother about it and all that kind of stuff.

Erik: Right. So it's kind of a, you know, the easy button for storage and what we are. What we believe at Vadura is that those two worlds are going to come together as this becomes more mature and more a part of our everyday computing life, right? So it's like, you don't want to have this kind of crazy, really hard to use storage on one side, right?

Erik: You want your really easy to use storage, but you want Fast, right? And that's the bet that we've made is that we believe that people want the performance of the fast file systems, but they want the easy use and in particular, the reliability of the enterprise world. And our product very directly brings those two things together into a single package.

Mehmet: Absolutely. Now I'm going to little bit shift gear and, you know, focus on your role as you know, uh, in product management, right? So [00:28:00] maybe it's gonna be kind of a stuffed question and sorry for that. But, um, so two things, Erik, the first one is, especially in the realm of, you know, machine learning AI, like how How easy or hard it is to stay, you know, from product management perspective up to speed with everything that's, you know, around you.

Mehmet: And the second thing, if we think about also, you know, The infrastructure part. Uh, some sometimes, you know, do you feel also like it's getting a lot off, uh, players in that space and you need really to do something special, both on the technical and commercial side to do the differentiation. So how do you You know, tackle these two points if you want.

Erik: That is kind of a tough question. Well, so I'll tell you, so if I, if I kind of go up to 30, 000 feet and I'd look at my [00:29:00] career, I've almost, you know, I've been in product management for great majority of it. Um, I would say at least for me, it's relatively easy to be right about what's going to happen. It's darn near impossible to be right about when it's going to happen, right?

Erik: So it's like the, the technology curves. If you pay close enough attention to the core technology, you can, you can see what's, what's happening, right? Because, you know, there's a relatively long timeline between a fundamental innovation and a product, you know, sometimes it's decades. Um, and so you can see it coming.

Erik: The thing that I've just continually been astonished by is how fast or slow some things happen. So I remember when I was in AMD, you know, we were predicting the demise of PCs, right? Oh, they're going down. And, you know, it turns out they've been going down in volume for, 20 20 years since then, right? But you know, they haven't they haven't died.

Erik: They went slow. I look at AI. It's exactly opposite You know, I remember looking at this Landscape a year ago. I'm thinking okay. Well, I think this is gonna happen and this is gonna [00:30:00] happen It'll probably happen over this time frame. But now I look a year later I'm like, that's exactly what happened, but it happened in one year instead of three.

Erik: So that's really been the The watchword for me is the speed of which the machine learning world is moving is just different, faster than anything I've ever seen in my whole life, but I've never seen a technology move at this pace. And that's the challenging part is just to keep up with it because of how quickly it's going.

Mehmet: Absolutely. And I remember when you know, we did the call before we do the actual recording today. This is one of the things that I mentioned to you. So even for me, it's challenging because sometimes, you know, we could have been discussing something today. And then by the time the episode is out, like, completely something new would come out and you know, like it become like kind of obsolete.

Mehmet: It's happened to me two times last year. So it's, it's really like a little bit even challenging for someone not into that [00:31:00] space. Now I want to talk a little bit about, you know, transitioning from, you know, being an engineer to product manager, uh, Erik, right. So, um, as an engineer, and I know from myself, I used to be an engineer myself.

Mehmet: So. You always, you know, have the itch to, to solve problems, right? So, so, but you do it on the technical level. What I get to know also that once you become a product manager and you are in the PM space, right? So you focus more on different aspects. So you still solve problems, but you focus more on like customer needs.

Mehmet: Uh, how they're going to use the product. So this shifting or this transitioning, um, from being an engineer to, to, to the field of product management, um, what it takes and. Is it like something, you know, everyone you think can do it, or there is [00:32:00] like some specific skills I need if I am an engineer today, I need to have to do this transition in my career.

Erik: Well, you know, that's a good question. I, uh, I'll tell you a story about how I became a product manager. So I was, I was young. I got out of college relatively, relatively young. Uh, start, you know, started, started early, got out fast. And, uh, and I had this idealistic, I remember telling myself at one time, I'll only use Excel to express my ideas.

Erik: I will never make a PowerPoint presentation in my life. And of course, all I've done for 20 years is presentations, but I was an engineer and I had a great, I had a great manager and I was talking to him one time. And I said, you know, these marketing guys, these product managers, they don't, they don't understand the technology.

Erik: And they did, as I recall, they'd come up with some roadmap and I was like, That's not going to work at all. Like the tech, it's not, you know, I mean, I, it was just obvious to me, it was a bad, it was not a good direction. And if you knew, you know, you knew where the technology was going, you just knew it was a bad strategy.

Erik: And I said, you know, they really need to do this. And you know, they're not thinking [00:33:00] about, you know, this new entrant that of technology that we have, blah, blah, blah. And I'm going off and he says, well, you know, I'll tell you what, kid, if you think you can do a better job, why don't you go do a better job?

Erik: Right. And that was, I don't know, almost 30 years ago. I've been doing it ever since. And that's the key for me. Has been, you have to understand three things equally well. One is you have to understand the technology and a lot of people forget about that, right? You just have to, you have to mess around with the technology.

Erik: You have to be a hobbyist. You have to be interested. You have to read all the, all the interesting news, right? You really have to pay attention because. The threat of new entrance, new technology is ultimately what's going to drive innovation in any industry. Right? So that's number one. Number two, and this was the hardest for me to learn, but the most important was talk to customers, right?

Erik: And I had another, I had another manager some years later who, you know, he told me, he says, Erik, just keep going out to the field, keep talking to customers because. If you just sit in the factory, if you just sit in corporate and, you know, [00:34:00] kind of eat your own dog food, you'll forget the pain they're going through, you'll forget the problems they're trying to solve, and you won't be very smart about about solving their problems.

Erik: And, you know, it's that, you know, you've got to talk to customers. And, of course, you have to talk to customers and you have to think about what they say, because, you know, if you say, Hey, you know, Mr. Customer, what should I build? Well, they're not the technology experts, so they're almost. Always going to say there'll be something 20 percent faster, 20 percent cheaper than what you're selling today.

Erik: And, you know, early in my career, I went off and did that. And then I'd come up and I, here's the thing. And you asked for like, that's true. We did ask for that, but then your competitor came out with this new technology and it's better. I'm really sorry about that instead. And so you've got to listen to the, you know, listen to them.

Erik: You got to think about what they think. They say it's very important, but also you have to think about what the technology is going to do, and that's the position that you have that's different, right? You're the expert in your field, right? You're not the expert in you like you have a customer that we have customers in the medical field.

Erik: I was just out at. And, you know, they're they're worried about the new technology in their world, and they're not [00:35:00] thinking about storage technology. So if I'm not thinking about that, nobody's thinking about. So, you know, understand the technology, understand what your customers real needs are and how to address them.

Erik: And then number three, understand what's realistic. In engineering and operational execution, right? I've seen a lot of product managers design the most beautiful, elegant products in the world, but nobody can build them. Uh, they can't design them and they can't get them can't build them once they are designed and, and they never get out to market.

Erik: So just a healthy dose of like operational. realism, operational, uh, awareness is, is very important. Those, those are, to me, those are the three legs of the stool.

Mehmet: To me, Erik, like it seems, uh, you know, I, I told you also before, like, I have this special also compassion for, uh, uh, people in product management, uh, because, you know, during my career, I had to deal a lot with them and they were super helpful for me and, you know, they were, you know, kind of guide to me.

Mehmet: Um, [00:36:00] so. What you mentioned now, you know, this balance between understanding, of course, the customer pain points and then understanding what's buildable. And then, you know, the engineering part also as well, and then making sure that you can market it in, in, in a proper way. Of course, like you don't do the marketing yourself, but you need to take this message.

Mehmet: So this communication with the marketing people, because part of, I know this for a fact is that you build the product, people they want, and then you design it in the best way possible. Of course, it's like a continuous process, but when it comes to the marketing part and, you know, of course, later giving it to the field sales people.

Mehmet: So again, shifting from engineering background, like how, how was this experience from your perspective, Erik?

Erik: You know, it's, um, It's, it's really, it's really interesting, you know, all the engineers think I'm a marketing guy, all the marketing guys think I'm an engineer, every [00:37:00] company I've ever, I've ever worked at, right?

Erik: But, you know, marketing in particular is such a delicate art, you know, people, um, I think there's, especially in technology companies, there's a tendency to give a lot of respect for the engineers and not a lot of respect to them. The fact is that it's, it's a very artful job, you know, and you don't know if you did well for probably a year.

Erik: After you did something right? I mean, the amount of time, um, is, is long. So you really have to be super, just thoughtful and careful about what you go do. But ultimately, marketing, the way I look at marketing is it's a, it's a delightful force multiplier, right? So if you put, you know, 10% of your, of your sales budget into marketing, 5%, 10%, you'll get something like a 10 x.

Erik: Multiplier on that investment, right? So ultimately good marketing, um, does two things, right? So I like that marketing There's there's two there's two two functions of marketing one is to Position and promote the products that you have in the [00:38:00] best way and number two is to pay enough attention to what's happening in industry To give feedback to the technologists early enough where they can do something about the information, right?

Erik: So there's actually an interesting two way Um, two way function of marketing that, you know, a lot of people overlook the second one, but very, very important,

Mehmet: right? Um, so, you know, we are almost like coming to an end Erik with you today. Uh, if, if you can. Want to give us with like final words of wisdom if you want to share anything, you know, that maybe we we didn't cover like just in a couple of minutes with with the audience.

Mehmet: What would that be? You know, I've had it

Erik: had an opportunity in my career to do a wide range of stuff, right? It's been just amazing. And I say the The thing that is the most fun and what I'm doing now is, you know, I spent, I spent essentially 25 years as a device manager and very challenging, very rewarding, [00:39:00] the companies I was with were very good to me, had a great career in it, but ultimately what you're doing is you're, you know, you're servicing solution providers.

Erik: And one of the things that I realized after some time of doing it is like, You get to see what everybody else is doing, right? And not not like precisely what they're doing, but you get to see all their practices, right? It's like, for example, when I saw these software companies coming in and saying, you know, our software just runs on any hardware, you know, off the commercial, off the shelf commodity hardware.

Erik: Well, you know, none of them succeeded because. Every time somebody tried to run the software, something would go wrong and then they'd have to use engineering to fix it. I was like, that's not going to work, right? And so, you know, you look at that, that's not a great, uh, that's not a great business model.

Erik: And, you know, so we figured out at Vadura, well, hey, we're going to create an appliance, but financially we're just going to be a software company. So customer gets the best of both worlds. 25 years of making devices, compute and storage has taught me. There are best practices in being a solution provider and the delightful journey that we've got at Vdura is we're able to go in and just put all those best practices in place and build a [00:40:00] great product for this brave new world that we're seeing.

Mehmet: That's fantastic. Final thing, Erik, where people can find more about Vdura and maybe get in touch with you.

Erik: Oh, go to www. Vdura. com, right? Like on my shirt. And, um, and come and see us at Super Compute. We're going to have a, we're going to have a big presence at Super Compute. We're super excited in November.

Mehmet: That's fantastic. Uh, Erik, really, I enjoyed the conversation. Like, uh, you have, uh, you know, a treasure with you. I mean, because you, you have seen a lot, uh, in your career, you know, and you know, you enlightened us a lot around, you know, everything related to, uh, AI storage systems, parallel computing, HPC. So we discussed a lot with you today.

Mehmet: So thank you very much for taking the time and for the audience, don't worry. The link will be in the, uh, in the show notes. If you are listening to us on any of your favorite podcasting [00:41:00] platforms, and if you are watching this on YouTube, you can also see it in the description. And this is usually how I end my episodes.

Mehmet: So if you just discovered this podcast by luck, thank you for passing by. I hope you enjoyed it. If you did so please subscribe. And share it with your friends and colleagues. And if you are one of the people who keeps coming and send me their comments and their suggestions, please keep doing so. I read all of them and I take all your comments into consideration.

Mehmet: Thank you very much for tuning in. We'll be again in a new episode very soon. Thank you. Bye bye.