Some of the topics that Matt covered include:
- Starting a company called CyberLabs and building bespoke rigs for gamers
- Public cloud versus private cloud
- The notion that organizations should be taking control of their systems and data
- Cybersecurity, data protection, and homomorphic encryption
- Material Science and the Future of storage technology
Some other titles we considered for this episode, but ultimately rejected:
- At That Point, I Knew Tech Had Me
- The Integration of the Software Layer and the Hardware Layer
- I Was Interested In Making Things Go Fast
- Storage In the Cloud Is Like a Fisher-Price Toy
- Latency Is Always a Challenge
Read the best-effort transcript
Read the best-effort transcript below (This technology is still not as good as they say it is…):
Michael Waitze 0:05
Okay, now we are actually officially on. Hi, this is Michael Waitze. And welcome back to the Asia Tech Podcast. We have Matthew Oostveen with us today the Chief Technology Officer, CTO of Pure Storage and a VP for Asia Pacific and Japan. Matt, it’s great to have you on the show.
Matthew Oostveen 0:23
Michael, thank you for having me.
Michael Waitze 0:25
It’s great to have you here. I love the fact that your headphones match the stripes on your T shirt. Like I can’t accomplish this. I don’t know how you do it. But it looks kind of cool. Anyway, it’s great to have you here.
Matthew Oostveen 0:35
It’s great to be here. Appreciate the time,
Michael Waitze 0:37
it is my pleasure. Look, I like to get people’s backgrounds, right, because I have this theory that you can’t be great at something unless you care about it. I always find that there’s something in somebody’s background that gives a hint as to why they do what they do. So let’s get some of your background for some context. Yes.
Matthew Oostveen 0:53
Yeah, absolutely. Let’s do that. The usual thing, because it’s almost job interview mode, isn’t it? You know, tell me about your career. But I don’t think that’s really answering the question. Because what I want to do is go back way, way, way further in time to when I was studying. And there was a period of, of transformation that was occurring in it in the mid 90s. And there was this new invention, and it was the 3d video card video cards right? Before then things were pretty simple on the PC. And then these video accelerators came along. And we saw a really good opportunity, I saw a really good opportunity. When these cards were coming out to start hand building and engineering systems, PCs, gaming rigs, if you will, and starting a company that was really focused on that. So I was 20 years old, my father had just retired from his butcher shop. So I had this kind of empty space that we cleared out. And we started a company called Cyber labs, over the course of you know, what was probably a short amount of time being a 20 year old and the attention span that you’ve got at that age. I constructed what what we were calling engineered PC. So it was really focused on on the gaming market, and really focused on getting the highest frame rates that were possible. And of course, this is, you know, almost 30 years ago, I think 9590 95. What I found was quite interesting, though, was this was this was a gateway, it was true that customers really didn’t want this type of product. It was it was nascent, but it was growing really fast. But the technical market started to get involved as well, they were saying, Well, you know, we want these 3d workstations for you know, CAD CAM applications, we want to, you know, the grant that you’re putting together from, from the systems, it was almost like Yum, char it right, you know, you’re picking and choosing the various parts and putting together something that was absolutely screaming, and then taking a really strong approach towards quality and output. So focusing really, really strongly on ensuring the systems were burnt in ensuring the systems were performing, you know, providing everyone who bought one of these systems with, you know, benchmark results of how it was tested. So that was my DNA. I probably spent maybe a year year and a half before moving on to other things, which was joining the IT industry. But at that point, you know, I knew that that Tech had me,
Michael Waitze 3:23
there are three things here that are really interesting. I mean, the first thing is entrepreneurialism, right. I mean, you see your dad running his own shop, and you just think okay, that space is now empty. What can I do with it? And in what am I really interested in? Sure there’s a point in time, right? Where these cards come out these voodoo cards, and I love this idea of kind of looking around and going everyone’s going to need this, but no one’s going to know how to do it. And I think this is going to be thematic, right? And then you figure, I need to do three things here, I need to build it, I need to deliver it. And then I need to service it, right. Because in a way the building quality is really important. But it’s kind of commoditized. But getting that delivery and that service to people. That’s where the differentiation is, at least for me, right? And once you get into the IT business at scale, the hardware and we’ll talk about hardware in a second right can be super duper important. But it’s like what you provide with that hardware that ends up being really important you learn this when you’re 20. Like do I have this wrong? Or or or my kinda right?
Matthew Oostveen 4:21
No, that’s absolutely the way that we saw it. And it was around that the integration between what was happening on the software layer and the hardware layer. So these machines were tuned for what the marketplace wanted, because there’s no point in having a 3d card if you’re not going to apply it towards, you know, the games that were available at the time. Let’s think back to the games at the time. By the way, we were just coming through the Doom era. And we were coming into id Software’s next big iteration, which was quake and quake was such a groundbreaking game at the time and it really did test your system to the max and this was also the area where then people were able to first link together systems in game. So we didn’t have the internet per se, it was very early days. But what we did have was, you know, IP addresses, you know, local land parties, the ability to link together PCs. And that’s where the first real competitive gaming world started to emerge from from, you know, I think, from Doom and from quake. So, the consumers were saying, Alright, well get me a rig that’s going to, you know, help me be competitive, I want to win these contests. And I want to be fast, I want the fastest framerate, and I want to be able to knock off the fastest game is.
Michael Waitze 5:37
So in 1995, I was at Morgan Stanley. And I remember when they came around to our desks and says, and said, Look, we’re going to install Local Area Networks back then we had, I think they were the smaller floppy disks. So I don’t even think they gave us access to the CD ROM drives, right? Because it was just too dangerous. And it confused people. But I remember thinking back then, okay, if it’s a LAN now, locally, then it’s going to be global, at some point, that means I’m gonna be able to share files with the guys and gals in London and in New York without actually having to email them and everything is just going to be so much easier to control. Did you see that then as well, right? Because you can see it happening as it’s going along. And you’re like, wait a second, what does this mean? what’s the implication of this, right? Because this gets really important as we get 2025 years later, because the technology world back then was changing really fast. It felt like it was changing really fast. But it’s changing even faster today. But I’m curious how you felt back then. And then we can transfer that to what’s going on today?
Matthew Oostveen 6:35
Yeah, you know, it’s funny, because I saw what you’re describing, but I saw it through a different lens. And at the time, as an entrepreneur, I knew that selling you know, ones and twos PCs was was fine. But if you really wanted to, you know, make more money than you need to sell at scale. And this is where that concept of local area networking came in, because organizations that were looking to digitize, if you will, their offices, you know, moved from that sort of physical realm and into this new digital realm. They were doing it in one big jump every single time. So that meant, you know, you had 10 workers or 20 workers, it meant, you know, 20 desktops, it meant then you needed to move to land. So I was very interested in networking very early. Because at that point, I realized that there was obviously larger opportunities available for, you know, from a revenue perspective. And then the next logical jump, of course, is, you know, going from, like you said, from lands to lands and board.
Michael Waitze 7:32
How do you go from being an entrepreneur, to then making the decision to say, I think there’s a lot more for me to do working at a bigger company. And again, I want to give you perspective on this too, because I want to tell you a story that was shared with me when I was still back in the financial services industry with a tech with a guy in the IT department, I said to him, you’re so good at this, why don’t you go out and start your own company. And he said to me without even thinking, and I thought it was kind of cool, actually. Right. He said, big banks have big problems. And it guys really want to solve big problems. And that’s why I’m here. Because I could go out and start my own company, and I maybe I make more money, maybe I make the same amount of money, maybe I can have more autonomy or whatever. But at the end of the day, I love solving big problems. And these are really big problems.
Matthew Oostveen 8:22
Yeah, that’s a good take. And I think that’s, there’s something to that, isn’t there? Yeah, mine isn’t too dissimilar. I wanted to I mean, I was interested in infrastructure, and I was interested in making things go fast, right? You know, Ricky, Bobby style, you know, how fast can I make a system? How fast can can we get, you know, supercomputer to crunch, for example. So that’s not something you’re going to do out of your father’s, you know, converted butcher shop. That’s where you need some some big brands behind you. And obviously, you know, the biggest supercomputing brand, arguably, at the time, I mean, you know, some people might argue Cray, but as we were moving into the later 90s, and it was time to, it’s time to use the degree, it was time to actually apply. And that was where IBM was, was, you know, a great starting point, because everyone gives you this career advice that says, if you want to learn tech, and if you want to learn business, then you know, that’s the greatest training ground on the world, in the world at the time. Yeah.
Michael Waitze 9:18
But I mean, it really was, and you brought up Cray, right? A lot of people that are listening to this may or may not remember when Cray supercomputer was the thing, but you could see it happen. And I think for guys like we are that we’re paying attention to what’s happening in the tech world. You could see them slowly but surely becoming less and less relevant. Because the ability to parallel process with just regular off the shelf PCs, whether it was from Dell or IBM or Compaq at the time, just was outstripping it right as Intel started making better and faster chips, and AMD was catching up with them as well. But when it changed, it now gives the ability to do things that we’re doing today. And I’m curious what you think too as you’ve gone through your entire career. When did the three putt and compute kind of changed so that you could do things either on prem or off prem. And it really wouldn’t make a difference.
Matthew Oostveen 10:09
Oh, see, that’s going to be so dependent. I don’t know if we’re even gotten there yet. My view is, if you want ultimate performance, and no compromises, then that’s not something that I would put into a public cloud. That’s something that I would keep, you know, I guess we’d say, on premises or in a private cloud, keep it close. I mean, latency is an issue. And there’s the concept we’re bringing into these concepts of data, gravity as well, you know, where’s that? Where’s the data stored? Where is the processing done, we want to make sure that the applications and the data, and the processing is all happening in a single location. If you’re creating a lot of information on premises, and it could be for a supercomputing application, for example, then the notion of putting that up somewhere else, and then bringing it back down again, you know, strikes me is as inefficient.
Michael Waitze 10:57
Yeah, look, I mean, I went through the same thing with my session, I want to get back to you said something just a second ago. And I don’t want you to forget about it. If you have the application, the data and the analytics, all in the same place, just just remember that and I’m sure I got one of those things wrong. But just remember that thought, even for someone like I am, who’s not running a gigantic business, I was storing all my stuff in the cloud. And just to process it, whether it was using Photoshop or Adobe Audition, it was just slow, like, it was latent. And I actually moved it out. And I did it for convenience, right? Because I was like, Look, if I’m at home, if I’m in my studio, or if I’m at a coffee shop, and I want to do some work, it’s better to have it in the cloud. And then I realized I was very inefficient, because the Photoshop was sitting on my desktop. But my data was in the cloud. So even for someone like I am, I could tell and I switched everything to my desktop machine. Can you go back and address this thing? We’re having all those things in the same place? Sorry.
Matthew Oostveen 11:54
I want to say something else as well. Because I think from a consumer perspective, we’ve gotten to a point where the cloud is obviously the manifestation of a business model, not something that’s there for convenience, it’s it strikes me is something that suits a corporation, and suits reporting to Wall Street because you’re showing reoccurring revenue. But from a consumer perspective, how much more useful is it to you know, to operate in the mechanisms that we’re talking about? It’s like using and I won’t name any brands, but you know, you use some, some printers? And some printers have as a service features in there is that, is that really necessary? Do I Do I really want to be able to consume in that way, because we all know that it’s going to be woefully expensive when compared to just doing it yourself. And once again, the printer becomes this physical manifestation of the organization’s business model, not something that’s they’re designed for consumer and to do something better.
Michael Waitze 12:49
It’s a super good point. It’s a super good point. So how do you balance that even in the even in the cloud stuff with what it costs me to do this? And with my level of efficiency, right? In other words, what is the business model there? How do you address that?
Matthew Oostveen 13:02
So we’ve gotten to a point where you’d say, Okay, five years ago, maybe seven years ago, people liked the public cloud because of convenience, because, you know, I didn’t need a large operations team, because I needed to move quickly, you know, that there was some really good reasons why you would want to do that. But also, the other reason was that the technology on premises hadn’t quite match what was happening in a public cloud. And I think that’s changed. And I don’t want to talk just about Pure Storage and the amount of engineering effort we’ve put into not just replicating but exceeding the capability of what a public cloud can do, but just bring those tools and and the the tools to deliver cloud like outcomes, if you will, so that people can do it themselves on premises on their own estates, and do so with far better performance with far lower cost. And without that, like I say that manifestation of a business model, it just feels like very, very designed for for the seller, not for the consumer yet feels
Michael Waitze 13:57
forced. It’s almost like it’s for them, not for me. Does that make sense? Yeah. Because the more you use it, the more you feel like you’re like, wait a second, that’s for them. What are you doing, then what’s like, give me the manifestation of this business model? Because then I want to dig deeper into exactly what’s happening.
Matthew Oostveen 14:21
Look, I think, firstly, is that the creation of these tools? It’s fundamentally important, but also the the notion that organizations should be taking control of their systems of their data. Were in a an era of a hyper awareness around security at the moment. In fact, it’s impossible now not to have discussions around what’s occurring from a security perspective. And the level of security that is available in the cloud. I mean, of course, it’s good, but it’s not to the same level that you’re going to get from, say, for example, an on premises storage array. The way that I like to describe this to people is to say You know, storage in the cloud is like a fisher price toy. You know, you almost want to shake it and it deletes and, you know, there’s not a lot to it. And you look at the rich features and functionality of these tune systems that you’re able to acquire now on premises and then with, you know, coupled with the engineering prowess that we’ve got within injuries or organizations, it becomes a far more attractive proposition for these organizations to pursue.
Michael Waitze 15:29
So are you familiar with a guy named David Heinemeier? Hansson, he runs Basecamp 37 signals and hate mail? Are you familiar with him at all?
Matthew Oostveen 15:36
No, I’m not tell me about it.
Michael Waitze 15:38
And to be fair, I wasn’t familiar with him either. But I spent a lot of time reading, right. And I just tried to understand what’s happening in the entire tech space. And I’ll I’ll read a bunch of stuff. And then also try to find out if this person is credible or not. Anyway, base camp, people use it 37 signals people know about this. And hey, mail is a thing that people you know, mail, something that people use. And he was he did a whole bunch of posts, sort of at the end of 2022, and into 2023, about just what you’re saying. It was way more efficient and way more cost effective for him to run this stuff internally than it was to run it externally. And I feel like there’s a change taking place with this, like, do you see that happening as well, where people are, don’t want to shake that Fisher Price toy, but they want to have more control over it? Is this some something that you’re seeing as well, when you go talk to your potential clients?
Matthew Oostveen 16:24
So I’ve got a region I cover Asia Pacific in Japan? So that’s everything from the timezones of India 30, New Zealand? Yeah, and it’s arguably, yeah, it’s big. And it’s nuanced, because you know, what’s working in New Zealand is gonna be really different from Japan, and the uptake in India is different from Korea, and so on, and so forth. So it’s a tale of looking down into the countries. But if we look at a country like Australia, or New Zealand, I’ll put them into a basket, what we saw a number of years ago was a really high uptake of cloud technology, which is not surprising, because those two countries have traditionally been very strong in the new uptake of attack when it comes through. Yeah, and there’s actually is almost another podcast of why this is the case, there was a book written about this, called the tyranny of distance, since the fact that, you know, these two countries are so far away from the rest of the world, that they’ve always looked towards technology to be able to do it themselves. But that aside, if we go to a country, like Japan, for example, they’re far more reserved when it comes to the uptake of a new tech, surprisingly. And so their cloud story is really different from the Australian cloud story, Australia is very advanced, Japan, still easing their way into this. So there’s a lot more to go from, you know, a North Asian perspective. But if we look in Australia, we look in New Zealand, we can see a different story being told we’re now seeing data coming back on premises, we’re now seeing a far more nuanced approach to where an application should reside, or a data should reside, based on those things we were talking about before, Michael, you know, cost latency performance.
Michael Waitze 18:00
We were thinking about this years ago, when I was working at Deutsche securities, right, and I was working on a portfolio trading desk. And that type of equity trading desk is uniquely global, right? Because if you’re running a global stock portfolio, you have stocks in the US the stocks in Europe and stocks in Asia, and the US would manage some of the US stocks, we would do some of it in Asia, and the European guys and gals would do some there. But we shared all this data. And we spent all this time trying to figure out how to remove that latency. Right? And we had all these arguments of Do we have one central database where we can all access the same thing. But what if that goes down? What’s the point? You know, single point of failure? How would we do that today? To make it the most efficient?
Matthew Oostveen 18:44
Oh, how would we do that today? I I think centralization is it’s a panacea, isn’t it? And you’re always gonna deal. And latency is always a challenge that you’re trying to overcome. And there’s almost an arms race isn’t there between the financial services organizations, particularly the high frequency traders to be able to reduce that and have some clarification? Correct. But there’s another side to this as well, which is around the productivity of the workforce and the utilization of of each of the workers because we’re now at a point where we’re watching salaries rising, we’re in a world where we’ve got rising inflation, the cost of goods is, is going up quite considerably. So CIOs and IT leaders are looking really seriously about how to not spend more on tech. Well actually not spend more on people but spend more on tech to assuage the people investments that they’ve already made. So all those things I was just saying about cloud, and how we’re seeing data coming from, you know, public cloud onto on premises. Well, I am also observing the opposite occurring when it comes to projects that are based on improving the productivity of the workforce. So that’s where we’re seeing centralized repositories in Single locations where a timezone of workers are able to, you know, simultaneously access the same data files and be able to work on that, because that improves collaboration and improves output of each of the workers.
Michael Waitze 20:15
There’s so much to talk about there. So I don’t think you can get away from productivity without talking about Chet GPT. At some level, I just don’t think it’s possible, right? I mean, artificial intelligence. It’s weird, right? Like, if we’d done this podcast six months ago, I wouldn’t have brought it up. But the whole world would have been talking about the metaverse and everything would have been purple. And I don’t understand why it’s purple. And we would have been talking about Bitcoin. Right, right, and get all of it. But but to be fair, that’s what we would have been talking about. But AI and what chat GPT is doing is something that’s been working on for at least 50 years, right? And I think we’re coming back to this point, I focus on this a lot where the compute and the connectivity is so ubiquitous now that it makes this type of thing possible. How does that, but there but there are multiple sides of this, right? So the chat up to the open AI stuff can make workers more productive. But it also talks to something else you brought up before, which is this data security? Because you have to balance these things. Right? So how do you how do you do that? How do you use the AI, but still feel like all of your stuff is secure as well?
Matthew Oostveen 21:20
Well, I can speak about this from from our own experience, rather than the advice that I would give to customers. So you know, from a pure perspective, obviously, we build our own software, we build our own systems, but most of the development and engineering that takes place at Pure Storage is is software driven. And we want to be really careful from an office, the CTO perspective to ensure that the workforce understands what they’re doing when they tap into GPT. To help with coding, you know, Can you can you make the script more efficient? For example, the question is, should be asked once that’s, you know, handed back to the user who owns that, who owns that IP? Yeah. Is it open AI? Because we would argue that it’s arguably not pure storage, you know, where did the what was the origin of that code? You know, was there open source origins, for example? And of course, there’s very big implications for commercial software construct if you’ve got open source software within there, because it would invalidate much of the contracts that you’re putting in place. Yeah, the point is that we’re walking very carefully into a legal minefield here, and people need to be quite cautious. I’ll give you another example of partner in a in a big for call it that. What’s their policy around the utilization of GPT? Well, you know, it’s outright banned. But if I talk to a consultant from a smaller organization will partake, what they’re using the chat GPT for is to replace the job of a junior. So in other words, Mr. Junior, could you please research, you know, there’s this small piece of information that I need, you know, run away, come back, you know, 48 hours later, I’m going to review your work, and then I’m going to incorporate it into the larger consulting project. Now, I can do that in 20 minutes. So I’m going to treat GPT like a junior, I’m going to check its work. I’m going to validate some of the claims that it’s making. And then I’m going to incorporate that inside that the large project now that seems like a smarter utilization of GPT. Because what are we doing? We’re assuaging knowledge workers and bolstering their capability not replacing them.
Michael Waitze 23:29
Yeah. I mean, look, this is this is a theme that I’ve been talking about for years, right? Because we saw this again, back when I was at, back when I was at Goldman, when Goldman Sachs acquired I think, was hauled tradings automated trading systems back in the late 90s, early 2000s. A lot of buy side, a lot of excuse me, sell side traders were concerned that they were going to lose their jobs, because algorithmic trading was going to replace their ability to handle stuff. Now what we did was we took a business that was making 8 million bucks a year and turn it into 135 by applying this technology to it, because it allowed us to scale in a way that wasn’t possible before. If you use so we’ve seen this happen. This is 20 years ago, 25 years ago, the same thing is gonna happen. But to be fair, we didn’t just let it ride. And in the same way, you can’t let open AI and chat GPT just ride like you said, you have to check that work. Right. And it’s, you say, go ahead.
Matthew Oostveen 24:20
Well, I was going to ask you a question. Go ahead. Do you know when you’re reading something that’s been created by chat GPT i Can you can you just detect that tone and the way that it’s written something because I’ve had friends that have have sent me responses? I said, Come on, you didn’t write that mess. Okay, you got me that was Chuck GPT. Because you just know that it’s coming from there. And I don’t know how to describe this sixth sense, but But do you know what I mean?
Michael Waitze 24:44
I know exactly what you mean. And I feel like I can tell and I’m pretty sure that I’m right now I’m going to pick a number 95% of the time because 5% of the time I’m definitely going to be either wrong or tired or both. But because you know this too, when you watch somebody give an answer You can tell when their PR team has already fed it to them. You’re like, give me the real answer. You know, you can tell by the way they’re speaking, by the way, they’re saying and the writing of something that’s automated is very different. Here’s the other thing, too. And I’m curious about your opinion on this. If the writing or the code, right, because GPT can also write code, but if it’s based on stuff that’s already been written, right, then you’re shaking your head. So go ahead. Well, I
Matthew Oostveen 25:29
Well, I mean, it’s, it’s, it’s yes, it’s technically been written, I suppose. Because what we’re talking about is billions and billions of possible combinations of words that are somewhat meaningless. And, you know, tokens are ascribed to these various combinations to bring out the most relevant rent and accurate response that hasn’t been written before. I would argue that it hasn’t. It’s like saying, you know, because the dictionary exists. Therefore, you know, it’s really the basis for every book that’s been written well, you know, kind of technically, but
Michael Waitze 26:04
kind of, kind of, kind of I push back a little bit, because based on what I understand about the way it works, is it saying here’s all this stuff that’s been written? Not hear all these individual words? Right. So the dictionary analogy, I get it. Yeah, you could say that. And you could have said that for 1000 years. But today, you can tell when you read something, that it’s kind of been, what’s the right word, not stolen, but taken from something else, and then combined with a bunch of other stuff. And sometimes when you read it, you can tell like, it sounds like it’s right. But boy, the thing that wrote it doesn’t understand it at all.
Matthew Oostveen 26:39
Yeah, yeah. But Cheez Its confident, though. But let’s, let’s think about this, right? So what are the what are the use cases where this is going to be problematic for us in the immediate future, rather than the long term future, and this is where we’re starting to see a change in insecurity and the type of attacks that are occurring from a phishing perspective. You know, in the past, you got a phishing email. And it was telling you that the CEO of the company quickly needed you to WhatsApp them. So you could transfer some money, or whatever the Tall Story was, maybe maybe it was a letter from your bank. And it went into some detail, and you’d read this thing and you go, my bank doesn’t talk like that it doesn’t use that tone. And you know, the grammatical errors, of course, mean that you filter it out pretty quickly, chat GPT has just changed that. And the way that or the language I like to use to describe it is it’s bedside manner, it has a tone in the way that it communicates. It’s very simple. It’s similar to corporate speak, it’s very similar to the way that you would get an email from, you know, XYZ bank. And so when you read this, you kind of have to double check, to make sure that you’re not falling for something or clicking on the wrong link, etc. So we are seeing a rise in attacks in the last six months, that have definitely got an improvement in the quality of writing and the language that’s being used.
Michael Waitze 28:00
How hard does it get for companies, and even individuals to protect against this when it’s now way easier for somebody to write spyware? Ransomware, you know, have hack attacks and stuff like that, because you can have chat GPT do it for you now, like before you actually have to know something? Yeah. How do you protect against all this stuff? And not just not just personally, but even like corporates protect all their data? Like, it’s just getting so complex? Yeah,
Matthew Oostveen 28:27
it is. And and I think that there’s, there’s a big change that’s taken place. And I’m not saying it’s driven by Chuck GPT. But just the security postures that we saw last year are different from the ones that we see this year. In last year, we were having discussions about this paradigm of us versus them. That’s what security was about, you know, wanted to keep bad guys out. And yeah, that’s still true. Of course, you want to do that. But there’s also a realization that, you know, they’re probably going to get in at some point. Yeah, take take that posture. So then the question is, how are you protecting the data? That’s, that’s sitting there an individual’s data? And how quickly are you able to bounce back if an attack occurs? So the conversation has shifted from keeping us away from them towards, you know, rapid recovery from an attack from the inevitable? You know, last year, Michael, we saw many headline attacks that were making front page news, you know, CEOs losing jobs, market, capitalizations of organizations dropping by, you know, billion plus dollars. So, you know, profound impact. You know, how do we go about trying to minimize that we’d bounce back quickly. We protect the data, we make sure that it’s encrypted at rest, encrypted in motion. You know, hopefully in the future, we’ll move to a world where we’ve got encrypted during processing as well. So that concept called homomorphic encryption, where we continue to increase the security posture. And by the way, when we move to that world of homomorphic encryption, that’s the world where it doesn’t really matter if a breach has occurred and data has been stolen, because it’s still be encrypted and it’s still locked from from prying eyes. So, you know, these are things we can look forward to. But at the moment we’re in an arms race against the the attackers and the crime gangs.
Michael Waitze 30:11
What is what is homomorphic
Matthew Oostveen 30:12
encryption? homomorphic encryption is the ability to be able to encrypt a dataset and then process that dataset without converting it, firstly to, to a plain file to an open file. So ordinarily, you might have your data encrypted, when it’s sitting inside a database, you would then move that data towards towards memory and towards the CPU. And at that point, it needs to be unencrypted, if you will. So it goes into a plain file where it becomes readable. And at that point, it’s vulnerable to attack. So homomorphic encryption means that the processing that takes place is done. So with encrypted data, now at the moment is really slow. It’s It’s painfully slow, it would be an enormous overhead, and an enormous burden for for an organization to be able to do that. And of course, the application would need to be rewritten. But this is, I think, something that we can expect to see. And, you know, a decade or two, yeah, I
Michael Waitze 31:11
mean, there’s like a, I was talking to this guy, actually, who runs a tech company in Australia around data security, and I just some struggling you, you heard me typing, I was just trying to look up his name, because we had him on the show as well. And I just can’t remember his name. So I’m sorry, very sorry to him. But I’ll put a link to it in the show notes. This idea of you need zero knowledge to actually understand what’s inside this data, but you have to understand how to access it. And he’s building this ability. I don’t know if it’s homomorphic encryption or not. But something that doesn’t require you like you said, to turn this into like a flat file or readable file to be able to actually act on it. Does quantum change this at all as well?
Matthew Oostveen 31:45
We’re not really sure. The certainly from a networking perspective, from a processing perspective, I would step back and say that sort of exceeds my my knowledge. But from a networking perspective, we know that quantum networking is inherently safe, because once the as you were saying, right at the start of the show, once the observation takes place, then the the link can be broken. So that it’s not possible to be able to eavesdrop if you will, on on data transfer. So that’s something that’s really interesting. And I think that that’ll have another profound effect on on the security posture.
Michael Waitze 32:23
So where do we go from here? Right? I mean, because the cloud is not going away. But this idea to access all your data on prem is not changing, like people are people doing work from home or really work from anywhere. And they’re gonna want to keep doing that. Like, I don’t think that’s going to go away. And you saw recently, Spain, I think, if I remember correctly, just announced a five year digital nomad visa. So people are going to stop they did, right. And I’m thinking about it like, I’ve never been to Spain.
Matthew Oostveen 32:51
Looks nice. Yeah, why not?
Michael Waitze 32:53
I don’t think I’d go to a visa, but but it would be nice to go to Barcelona once. But what’s the implication for storage and security and all this other stuff, if people are going to be able to work from anywhere? And remember, these are people from? These are also people from anywhere? Right? So you’re talking about guys and gals from Vietnam, working together with people from guys and gals from New Zealand, living in spit, like, what’s the implication of all this stuff?
Matthew Oostveen 33:21
Yeah, so it? That’s a really good question. Because where your data resides, and the type of access to that data matters to to answer the question. Yeah. So we’re moving towards a world where, you know, absolutely, the public cloud and centralized data repositories are going to be important, you know, think, and I’m talking more at a consumer level here. So think, cold storage, that doesn’t have a lot of access to the file types. And then there’s the data that we carry with us. So it’s your SSD drive in your laptop, or it could be the amount of storage that you’ve got within a phone? Well, geez, we use that quite a lot, don’t we? And that’s the high levels of IO there. So the next 20 years or so we’re going to see an incredible increase in the amount of storage capability that we’ve got. I know, from a pure perspective, if I look at the roadmap that we’ve got for our arrays, the density that we’re now able to put into what we call the DFM. So that is a direct flash module, it’s what holds the data in in one of these storage systems. The the amount of data that we’re able to fit within there is is growing faster than, you know, digital version of Moore’s law or storage version of really. So what that means is within a data center, provider, customers go down the path of choosing solid state flash technology that will be able to decrease their footprint, you know, unless they’re growing at some, you know, absurd amount of data growth, which you’d only see it in a very early phase startup anyway, but for most established organizations, you’re not going to get that that growth rate. So well, that’s a really big Positive, it’s a positive on a number of areas, it’s certainly positive from an environmental perspective, because you know, footprints are going to shrink within a datacenter. We’re going to use less power within a datacenter. And I mean, massive amounts less power by moving towards this solid state technology. But also, we’re going to have an incredible amount of storage available to us in our handheld devices, because we’ll soon see this technology rippled through into you know, your iPhones and your Android. How far along the
Michael Waitze 35:29
development curve, are we in solid state? And I’ll tell you why I’m asking. You know, I’ve been using spinning disks. First of all, I don’t know how old you are, I can probably guess if you said you did something 30 years ago, when you’re 20, you have to be around 50, late 40s, whatever, but, but I remember using like a Bernoulli disk. And that Bernoulli disk was like about as big it was bigger than an iPad and stickups is about 10 of them together. All right. And that was like that. If you had one at your desk, you were like one of the cool kids at Morgan Stanley, because it was a lot of storage. I don’t remember what it was was like five megabytes or 10 megabytes or something, right. But I remember spinning disks as well. And I’ll tell you this, I went out to buy a backup party, you can see me pointing over to the left, I was pulled up to buy a drive. And I was like I wanted to make sure that it was solid safe. But like how, because it was so much faster, and so much is used in a spinning hard drive. But how far along the development curve? Are we? And is there something coming in after that? I mean, you wouldn’t know. Right?
Matthew Oostveen 36:29
Yeah, so we have material science specialists within office of the CTO, wow, we are the only storage company on the planet enterprise storage company that buys directly from a fab plant from an ad plant, you know, everyone else buys even the big guys by one or two steps away. Because we’re building our own D FM’s these direct flash modules. So material science is really important to us. Yeah, I can share with you please, you know, what we think the future is going to look like from material science perspective and, and tell that story, you need to understand how NAND works and what a, you know what a flash module can hold. And to be super simple about this, I’ll be as simple as possible, because it can get complex really quickly. Think of you know that the bit where you store the bit of information as a cell, and inside that cell, you know, you’ve got two parts to it. There’s the positive and the negative in order to take one bit. Well, that’s called SLC. So that’s a single layer cell. And then you go to MLC, so that’s a multilayer cells. So that’s where we start to put two bits in there. And then you can go to TLC, which is at three. And then we’re now in a world where we’ve got this technology called Q LC. So it’s quite level. And this is why we’re seeing this hockey stick towards density, because we’re able to get incredible levels of density and performance out of out of these systems and high levels of efficiency as well, you know, uses far less power. But as we start to stuffing more cells, so bits into the cells, we start to negatively impact the performance. So there’s a trade off here.
Michael Waitze 38:13
Go ahead, no question. No, go ahead. I’m just thinking, I’m trying to visualize these cells, this these bits sitting inside these cells, right and in what did you call it? An SLC where it’s just a single cell, like inside of there just one bit, and it’s just by itself, and it’s looking around going? Okay, I’m just waiting for someone to access me 01 I get it. But if you’re getting up to quad now it’s getting crowded
Matthew Oostveen 38:35
with LEDs, and it don’t think of it as four. It’s more like 16.
Michael Waitze 38:39
Because I mean, I didn’t know what you meant by quiet, but it’s about power to at some level. Right.
Matthew Oostveen 38:44
Right. Right. So that’s typical approach from a computing perspective.
Michael Waitze 38:49
We’re second. Yeah. So two to the fourth power. 16. Is that would quad is correct. Got it?
Matthew Oostveen 38:55
So so if you were to look at a diagram, you would see I got it. I got all these elements of three actually
Michael Waitze 39:01
means eight then because it’s 248. Yeah,
Matthew Oostveen 39:04
correct. Oh, four, correct. And then then every time you’re reading and writing to one of these cells, you’re wearing that NAND out in the lives, the lifespan of it is probably say eight years, 10 years, seven years, it depends on on the utilization of the particular type of storage. But also, every time you read and every time you write up, that’s also influencing the amount of energy usage. So that needs you need to be very light in the way that you interact with this technology. So then the question is, is either one hour today, What’s tomorrow? You know, can we put five bits per cell in there and that’s something that we would call PLC. That’s a Penta level cell. PLC technology. Obviously, we’re playing with that in laboratories. We’re talking to the NAND plants about that. They’re talking to us about that as well. I wouldn’t hold my here’s my view. I don’t think that as it stands, it’s a technology that I would put inside an enterprise product in the next few years. I’ll tell you why. Firstly, is that the wear rate is very high on scale severs is going to degrade
Michael Waitze 40:13
really fast, we’re gonna have real data saving problems, but go ahead,
Matthew Oostveen 40:17
right. And the second thing is that you’re not really getting the performance game, you know, it starts to flatten out because this is a very busy cell after all. So this is where we’re starting to get into the layering of we’ll probably build more layers into TLC and more layers into Q LC. And that’s really going to be the main story, I don’t see NAND going anything further down the path of, of, of compression from PLC, but it doesn’t really finish the story and the narrative around what’s next from a solid state perspective.
Michael Waitze 40:51
Yeah, so it doesn’t at all, actually, you’ve mentioned energy a bunch of times, how important is it for a company like yours to control the energy usage? And I mean, not just from a cost perspective, but also from an environmental perspective? Because you you brought it up? Like, is this something people actively talk about and think about internally?
Matthew Oostveen 41:11
Yeah, not just internally. Mostly, this is customer driven justing, the the observation that I would make is maybe a couple of years ago, a year ago, two years ago, I would have conversations around ESG, with my C level counterparts at a financial services organization will be government departments. And I’m not saying they were platitudes that were exchanged back and forth. I mean, they were well intentioned, but there was never really a meat behind it. And what’s changed is, the C suite started to enable what I call the M suite, you know, the managers got involved. And you know, what’s going to happen when a manager is involved, Michael, they’re going to put a KPIs next to these three things. So now, we’re seeing individual contributors talking about ESG, because they’ve been tasked with reducing electricity consumption in a data center, for example, by 10%. Or if you’re in the UK, where prices of electricity have gone up, 600%, probably a lot more. So we’re seeing this coming together of you know, one wave being rising energy costs, and the other wave, which is, you know, reducing carbon dioxide emission. And that all points to a lot more dialogue around what organizations should be doing from apparent cooling perspective, which means the conversation has changed. Last year, it was about features and speeds and feeds and these things about storage today. Tell me about you know, how many watts per petabyte, for example, your storage array uses, you know, talk to me in that language, not the language of how faster a database performs.
Michael Waitze 42:44
Do you remember, it’s so interesting, but do you remember the feeling you had you said, when you were a 20 year old, right? When your dad retired, you had this open space in the butcher shop, and you started building these, and you said, engineered? I loved this terminology, engineered rigs, right. So it wasn’t like what everybody else was building, they were specifically engineered, and back then you were saying it was for gaming. As a 20 year old, you must have been super excited to like, run down with your buddies to build these things. Because it was kind of at the cutting edge. And even if it wasn’t cutting edge, it’s sort of felt like it to the 20 year old you I hear in your voice this kind of what I’m presuming is the same excitement. Do you feel the same way about what you’re doing today that you did then? Or is it even more exciting because the tech curve is just advancing so rapidly, and the stuff you’re working on? sounds so interesting.
Matthew Oostveen 43:33
Here’s what’s different. When I was 20 years old, I was at the whim of Intel, you know, as cork in the ocean, it depended on what they released. Fair enough. Now, now I’m part of it. And now I’m helping to build it. Now I’m seeing what the projects look like. Now I’m providing the feedback about what customers want here in Asia to ensure that we’re building the right tech from from an engineering perspective, so they feel far more like my own creations. Rather than how I use that term. Yamcha it you know, you pick a few things off the trolley and create your own lunch. Now, we’re genuinely building some things that are a game changing, and I can see what the roadmap looks like moving forward. I’m kind of blown away.
Michael Waitze 44:16
Okay, I think that’s the best way to add Matthew Oostveen the CTO, and a VP of Pure Storage for Asia Pacific in Japan. Thank you so much for doing that today. That was awesome.
Matthew Oostveen 44:27
You’re welcome, Michael. Lovely