 From around the globe, it's theCUBE, presenting the convergence of file and object, brought to you by Pure Storage. We're back with the convergence of file and object, a special program made possible by Pure Storage and co-created with theCUBE. So in this series, we're exploring that convergence between file and object storage. We're digging into the trends, the architectures, and some of the use cases for unified, fast file and object storage, UFFO. With me is Matt Burr, who's the vice president and general manager of FlashBlade at Pure Storage. Hello, Matt, how you doing? I'm doing great, morning Dave, how are you? Good, thank you. Hey, let's start with a little 101, kind of the basics. What is unified, fast, file and object? Yeah, so look, I mean, I think you got to start with first principles talking about the rise of unstructured data. So when we think about unstructured data, and you sort of think about the projections, 80% of data by 2025 is going to be unstructured data, whether that's machine generated data or AI and ML type workloads, you start to sort of see this, I don't want to say it's a boom, but it's sort of a renaissance for unstructured data, if you will, we move away from, what we've traditionally thought of as general purpose NAS and file shares to really things that focus on fast object, taking advantage of that three cloud native applications that need to integrate with applications onsite, AI workloads, ML workloads, tend to look to share data across multiple data sets, and you really need to have a platform that can deliver both highly performant and scalable fast file and object from one system. So talk a little bit more about some of the drivers that bring forth that need to unify file and object. Yeah, I mean, look, there's a real challenge in managing bespoke infrastructure or architectures around general purpose NAS and DAZ, et cetera. So if you think about how an architect sort of looks at an application, they might say, well, okay, I need to have fast DAZ storage proximal to the application, but that's going to require a tremendous amount of DAZ, which is a tremendous amount of drives, right? Hard drives are historically pretty unwieldy to manage because you're replacing them relatively consistently at a multi-paddabyte scale. So you start to look at things like the complexity of DAZ, you start to look at the complexity of general purpose NAS, and you start to just look at, quite frankly, something that a lot of people don't really want to talk about anymore, but actual data center space, right? Like consolidation matters, the ability to take something that's the size of a microwave like a modern flash blade or a modern, you know, UFO device, replaces something that might be the size of three or four or five refrigerators. So Matt, why is now the right time for this? I mean, for years, nobody really paid much attention to object. S3 obviously changed that course. Most of the world's data is still stored in file formats as you get there with NFS or SMB. Why is now the time to think about unifying object and file? Well, because we're moving to things like a contactless society, you know, the things that we're going to do are going to just require a tremendous amount of more compute power network and quite frankly, storage throughput. And, you know, I can give you two sort of real primary examples here, right? You know, warehouses are being, you know, taken over by robots, if you will. It's not a war, it's sort of a friendly advancement in, you know, how do I store a box in a warehouse? And, you know, we have a customer who focuses on large sort of big box distribution warehousing and, you know, a box that carried an object two weeks ago might have a different box size two weeks later. Well, that robot needs to know where the space is in the data center in order to put it, but also needs to be able to process, hey, I don't want to put the thing that I'm going to access the most in the back of the warehouse. I'm going to put that thing in the front of the warehouse. All of those types of data, you know, sort of real-time, you can think of the robot as almost an edge device, is processing in real-time unstructured data and its object, right? So it's sort of the emergence of these new types of workloads. And I give you the opposite example, the other end of the spectrum is ransomware, right? You know, today, we'll talk to customers and they'll say quite commonly, hey, if anybody can sell me a backup device, I need something that can restore quickly. If you had the ability to restore something in 270 terabytes an hour or 250 terabytes an hour, that's much faster when you're dealing with a ransomware attack, you want to get your data back quickly. You know, so I was going to ask you about that later, but since you brought it up, what is the right, I guess, called architecture for ransomware? And explain like how unified object and file would support, I mean, I get the fast recovery, but how would you recommend a customer go about architecting a ransomware proof system? Yeah, well, you know, with FlashBlade and with FlashArray, there's a natural feature called Safe Mode. And that Safe Mode actually protects the snapshots and the data from sort of being a part of the ransomware event. And so if you're in a type of ransomware situation like this, you're able to leverage Safe Mode and you say, okay, what happens in a ransomware attack is you can't get access to your data. And so, you know, the bad guy, the perpetrator, is basically saying, hey, I'm not going to give you access to your data until you pay me, you know, X in Bitcoin or whatever it might be, right? With Safe Mode, those snapshots are actually protected outside of the ransomware blast zone and you can bring back those snapshots because what's your alternative? If you're not doing something like that, your alternative is either to pay and unlock your data or you have to start restoring, excuse me, from tape or slow disk that could take you days or weeks to get your data back. So leveraging Safe Mode, you know, any of the FlashArray to FlashBlade product is a great way to go about architecting against ransomware. So I got to put my, I'm thinking like a customer now. So Safe Mode, so that's an immutable mode, right? Can't change the data. Is it, can an administrator go in and change that mode? Can he turn it off? Do I still need an air gap, for example? What would you recommend there? Yeah, so there are still, you know, sort of RBAC or Role-Based Access Control policies around who can access that Safe Mode and who can't. Right, okay. So anyway, subject for a different day. I want to actually bring up, if you don't object, a topic that I think used to be really front and center and now it's becoming front and center again. I mean, Wikibon just produced a research note forecasting the future of Flash and hard drives. And those of you who follow us know we've done this for quite some time and you can, if you could bring up the chart here. You could, and we see this happening again. It was originally we forecast the death of quote unquote high spin speed disk drives which is kind of an oxymoron. But you can see here on this chart, this hard disk had a magnificent journey but they peaked in volume in manufacturing volume in 2010. And the reason why that is so important is that volumes now are steadily dropping. You can see that. And we use Wright's law to explain why this is a problem. And Wright's law essentially says that as your cumulative manufacturing volume doubles your cost to manufacture declined by a constant percentage. I won't go too much detail on that but suffice it to say that flash volumes are growing very rapidly, HDD volumes aren't. And so flash because of consumer volumes can take advantage of Wright's law and that constant reduction. And that was really important for the next generation which is always more expensive to build. And so this kind of marks the beginning of the end. Matt, what do you think? What's the future hold for spinning disk in your view? Well, I can give you the answer on two levels. On a personal level, it's why I come to work every day. You know, the eradication or extinction of an inefficient thing. I like to say that inefficiency is the bane of my existence. And I think hard drives are largely inefficient. And I'm willing to accept the sort of longstanding argument that, you know, we've seen this transition in block, right? And we're starting to see it repeat itself in unstructured data. And I'm willing to accept the argument that cost is a vector here and it most certainly is, right? HDDs have been considerably cheaper than flash storage even to this day, up to this point, right? But we're starting to approach the point where you sort of reach a 3x sort of differentiator between the cost of an HDD and an SDD. And that really is that point in time when you begin to pick up a lot of volume and velocity. And so that tends to map directly to what you're seeing here, which is a slow decline which I think is gonna become even more rapid kind of probably starting around next year where you start to see SSDs, excuse me, SSDs, really replacing HDDs at a much more rapid clip, particularly on the unstructured data side and it's largely around cost. The workloads that we talked about, robots and warehouses or other types of advanced machine learning and artificial intelligence type applications and workflows, they require a degree of performance that a hard drive just can't deliver. We are seeing sort of the creative, innovative disruption of an entire industry up before our eyes. It's a fun thing to live through. Yeah, and we would agree. I mean, the premise there is it doesn't have to be less expensive. We think it will be by the second half or early second half of this decade. But even if it's, we think around a 3x Delta, the value of SSD relative to spinning disk is going to overwhelm just like with your laptop. It got to the point where you said, why would I ever have a spinning disk in my laptop? We see the same thing happening here. And so, and we're talking about raw capacity, put in compression and de-dup and everything else that you really can't do with spinning disk because of the performance issues you can do with flash. Okay, let's come back to UFFO. Can we dig into the challenges specifically that this solves for customers? Give me, give us some examples. Yeah, so, I mean, if we think about the examples, the robotic one, I think is the one that I think is the marker for kind of the modern side of what we see here. But what we're seeing from a trend perspective, which not everybody is deploying robots, right? There's many companies that aren't gonna be in the robotic business or even thinking about sort of future type oriented type things. But what they are doing is Greenfield applications are being built on object, generally not on file and not on block. And so, the rise of object as sort of the, sort of let's call it the next great protocol for modern workloads, right? This is that modern application coming to the forefront. And that could be anything from financial institutions right down through, we've even seen it in oil and gas. We're also seeing it across healthcare. So, as companies take the opportunity, as industries take this opportunity to modernize, they're modernizing not on things that are leveraging sort of archaic disk technology, they're really focusing on object, but they still have file workflows that they need to be able to support. And so having the ability to be able to deliver those things from one device in a capacity orientation or a performance orientation, while at the same time dramatically simplifying the overall administration of your environment, both physically and non-physically, is a key driver. So the great thing about object is it's simple. It's kind of a get put metaphor. It scales out, you know, because it's got metadata associated with the data and it's cheap. The drawback is you don't necessarily associate it with high performance and as well, most applications don't speak in that language. They speak in the language of file, you know, or as you mentioned block. So I see real opportunities here. If I have some data that's not necessarily frequently accessed, you know, every day, but yet I want to then end of quarter or whatever it is. I want to machine learning. I want to apply some AI to that data. I want to bring it in and then apply a file format because for performance reasons, is that right? Maybe you could unpack that a little bit. Yeah, so, you know, we see, I think you described it well, right? But I don't think object necessarily has to be slow. Nor does it have to be, you know, because when you think about, you brought up a good point with metadata, right? Being able to scale to a billions of object, being able to scale to billions of objects, excuse me, is of value, right? And I think people do traditionally associate object with slow, but it's not necessarily slow anymore, right? We did a sort of unofficial survey of our customers and our employee base. And when people described object, they thought of it as like law firms and storing a word doc, if you will. And that's just, you know, I think that there's a lack of understanding or a misnomer around what modern object has become and perform an object, particularly at scale when we're talking about billions of objects, you know, that's the next frontier, right? Is it at pace performance-wise with, you know, the other protocols? No, but it's making leaps and grounds. So you've talked a little bit more about some of the verticals that you see. I mean, I think when I think of financial services, I think transaction processing, but of course they have a lot of tons of unstructured data. Are there any patterns you're seeing by vertical market? We're, you know, we're not. That's the interesting thing. And you know, as a company with a block heritage or a block DNA, those patterns were pretty easy to spot, right? There were a certain number of databases that you really needed to support, Oracle, SQL, some Postgres work, et cetera, then kind of the modern databases around Cassandra and things like that. You knew that there were gonna be VMware environments, you know, you could sort of see the trends and where things were going. Unstructured data is such a broader horizontal thing, right? So, you know, inside of oil and gas, for example, you have, you know, you have specific applications and bespoke infrastructures for those applications, you know, inside of media entertainment, you have the same thing. The trend that we're seeing, the commonality that we're seeing is the modernization of, you know, object as a starting point for all of the, all of the net new workloads within those industry verticals, right? That's the most common request we see is what's your object roadmap? What's your, you know, what's your, what's your object strategy? You know, where do you think, where do you think object is going? So, there isn't any, you know, sort of, there's no, there's no path. It's really just kind of a wide open field in front of us with common requests across all industries. So the amazing thing about Pure, just as a kind of a little quasi, you know, armchair historian in the industry, Pure was really the only company in many, many years to be able to achieve escape velocity, break through a billion dollars. I mean, 3-par couldn't do it, Isilon couldn't do it, Compellent couldn't do it, I could go on, but Pure was able to achieve that as an independent company. And so you become a leader. You look at the Gartner Magic Quadrant, you're a leader in there. I mean, if you made it this far, you got to have some chops. And so, of course, it's very competitive. There are a number of other storage suppliers that have announced products that unify object and file. So I'm interested in how Pure differentiates, why Pure? It's a great question. And it's one that, you know, having been a long time Puritan, you know, I take pride in answering, and it's actually a really simple answer. It's business model, innovation, and technology, right? The technology that goes behind how we do what we do, right? And I don't mean the product, right? Innovation is product, but having a better support model, for example, or having on the business model side, you know, evergreen storage, right? Where we sort of look at your relationship to us as a subscription, right? We're gonna sort of take the thing that you've had and we're gonna modernize that thing in place over time, such that you're not re-buying that same, you know, terabyte or, you know, petabyte of storage that you've paid for over time. So, you know, sort of three legs of the stool that have made Pure clearly differentiated. I think the market has recognized that. You're right, it's hard to break through to a billion dollars. But I look forward to the day that, you know, we have two billion dollar products. And I think with, you know, that rise in unstructured data growing to 80% by 2025 and, you know, the massive transition that, you know, you guys have noted in your HDD slide, I think it's a huge opportunity for us on, you know, the other unstructured data side of the house. You know, the other thing I'd add, Matt, and I've talked to Cos about this is it's simplicity first. I've asked him, why don't you do this? Why don't you do that? And the answer is always the same. It's that adds complexity. And we put simplicity for the customer ahead of everything else. And I think that served you very, very well. What about the economics of unified file and object? I mean, if you're bringing additional value, presumably there's a cost to that, but there's got to be also a business case behind it. What kind of impact have you seen with customers? Yeah, I mean, look, I'll go back to something I mentioned earlier, which is just the reclamation of floor space and power and cooling, right? There's a, people want to search for kind of the sexier element, if you will, when it comes to looking at how you derive value from something, but the reality is if you're reducing your power consumption by a material percentage, power bills matter in big data centers. Customers typically are facing a paradigm of, well, I want to go to the cloud, but the clouds are not being more expensive than I thought it was going to be. Or I've figured out what I can use in the cloud. I thought it was going to be everything, but it's not going to be everything. So hybrids where we're landing, but I want to be out of the data center business and I don't want to have a team of 20 storage people to administer my storage. So there's sort of this very tangible value around, hey, if I could manage multiple petabytes with one full-time engineer, because the system to your and Kaz's point was radically simpler to administer, didn't require someone to be running around, swapping drives all the time, would that be a value? The answer is yes, 100% of the time, right? And then you start to look at, okay, all right, well, on the UFFO side, from a product perspective, hey, if I have to manage a bespoke environment for this application, if I have to manage a bespoke environment for this application and a bespoke environment for this application and a bespoke environment for this application, I'm managing four different things. And can I actually share data across those four different things? There's ways to share data, but most customers, it gets too complex. How do you even know what your gold.master copy is of data? If you have it in four different places or you try to have it in four different places and it's four different siloed infrastructures. So when you get to the sort of the side of, how do you measure value in UFFO, it's actually being able to have all of that data concentrated in one place so that you can share it from application to application. Got it. I'm interested, we have a couple minutes left. I'm interested in the update on FlashBlade, generally, but also I have a specific question. I mean, look, getting file right is hard enough. You just announced SMB support for FlashBlade. I'm interested in how that fits in. I think it's kind of obvious with file and object converging, but give us the update on FlashBlade and maybe you could address that specific question. Yeah, so look, I mean, we're tremendously excited about the growth of FlashBlade. We found workloads we never expected to find. The rapid restore workload was one that's actually brought to us from a customer, actually, and has become one of our top two, three, four workloads. So we're really happy with the trend we've seen in it and mapping back to thinking about HDDs and SSDs. We're well on a path to building a billion-dollar business here, so we're very excited about that. But to your point, you don't just snap your fingers and get there, right? We've learned that doing file and object is harder than block because there's more things that you have to go do. From one, you're basically focused on three protocols, SMB, NFS, and S3, not necessarily in that order. But to your point about SMB, we are on the path through to releasing SMB, full native SMB support in the system. That will allow us to service customers. We have a limitation with some customers today where they'll have an SMB portion of their NFS workflow. And we do great on the NFS side, but we didn't have the ability to plug into the SMB component of their workflow. So that's gonna open up a lot of opportunity for us on that front. And we continue to invest significantly across the board in areas like security, which has become more than just a hot button. Today, security has always been there, but it feels like it's blazing hot today. And so going through the next couple of years, we'll be looking at developing some pretty material security elements of the product as well. So well on a path to a billion dollars is the net on that. And we're fortunate to have SMB here. And we're looking forward to introducing that to those customers that have NFS workloads today with an SMB component. Yeah, nice tailwind, good TAM expansion strategy. Matt, thanks so much. We're out of time, but really appreciate you coming on the program. We appreciate you having us. And thanks much, Dave. Good to see you. All right, good to see you. And you're watching the convergence of file and object. Keep it right there. We'll be back with more right after this short break.