Transcript
I'm your host, Rob Luedemann. It is time to bring the orange, and I am pleased to have back in studio Chad Kenney, VP of Product Management, and we are doing launch episode number two all around the unified data plane stuff that you know and love, but something that you absolutely have to have in place to be successful with what we talked about on episode one with the intelligent control plane. Oh, yeah. Automation. What happens if you don't have that? It breaks down, right? Everything breaks down. Yeah. I think we take this for granted in many cases. We have just a killer platform built on an evergreen architecture. People can upgrade non-disruptively, and we can cover from archive level all the way up into the most performant in memory, AI. It's a good time to be here. We've got coverage pretty much on every possible use case, and with the intelligent control plane, get to manage all of it and govern it with the intents, which is kind of fun. Yeah. It's kind of mind-blowing when I look at the slide or the visual that shows all the different use cases along the bottom, because I used to make some of those when I was in my prior role, and we just had like a few dots. It's like, we're really good at databases, and we're really good at VM workloads. VDI was a big thing back then. VDI was huge. Nobody wants a bootstorm, but now we look across that, and truly, with the investments that we've made, always great with block, but in file capability, with object capability, and now having that span the entire portfolio, basically the entire platform, we're providing that power to customers and also making it so they don't have to worry about it anymore. You don't have to go spend time being a storage jockey trying to figure out what needs to go where, set up policy, and then go do something else that's more interesting. That abstraction that we built kind of from the very get-go, where we got rid of like the RAID groups, and we got rid of even the SSDs, in fact, we got rid of the FDL exists, but globally managed flash, we're now just abstracting it further. Now we're managing a fleet. It's like one system, but really many systems behind it. It's pretty fun, because I think once you start to think about it in that model, these new innovations that we're building into the data plane actually just start to add more and more opportunities for them to get value from it, as well as consolidate more workloads to it, and we have more hexagons, I guess. Yeah. Yeah. Remember back in the day in virtualization, I was laughing at this the other day, it was the I-O blender. Do you remember this? Yes, I remember the I-O. We had the little graphic. Yeah, that was fun. Like the little Y flash became kind of cool, because you had all these workloads running at the same time. But if you think about it, what we're doing now is it's across the data plane. We are blending so many different things, and we're getting the data plane to be able to land those things in the right location through the intelligent control plane. But also unshackling that old-school methodology of like one workload attached to one system. And again, I put my like server guy hat on. It was always like, well, this is our mission-critical thing. And the notion of tiering, right? I don't even know if we hear or talk about tiering much anymore, because you don't really have to worry about it. We can treat all workloads equally with the same level. But also, if you don't want to have that, great, we can move it around. It's about choice, right? It's about choice where you want to deploy, but also not having to worry about it at night. And defining the service level you want to deliver, and telling it to go do so. That's fun. Yeah, super fun. It's super exciting time. I'm super looking forward to getting into it, and I love that you hit kind of the protocol side of like, it doesn't matter. But also now we're also at a point where we can talk about truly edge-to-core-to-cloud, that extension, and we'll hit on some of those in this episode as well, about where we continue to innovate like out at the cloud layer, but also like into the edge. Super fun. Let's do it. Data keeps blowing up everywhere, right? So the stat we hit is like 175 zettabytes, and I love to always contrast this to my dad who's an old storage guy, and at the company, at the blue company he worked at with three letters a long time ago, like their global target in 1989, just to blow people's minds, was to sell one gig worth of storage to the entire planet. And now we're just exponentially blowing that away. It's crazy what all the new apps and everything is driving to explode data, and our customers and our prospects are having to contend with that, and we're here for it. It's gone from that very structured environment, which is very much still important, to this now series of data that is going all over the place, from videos to images to log-based data to everything you can possibly imagine. So I think when you think about a unified data plan, you've got to think about that you need to be able to land everything, right? Like there's really no room to maneuver effectively on not having a place to land certain things. And so the big benefit here is that we need to continue to expand both down into cheaper and more cost-effective solutions, which is what we've been doing with cyber and archived tiers in the past. We also need to be able to move further up the stack. We did a fair bit of it last year with continually evolving the platforms to the next generation, get new features, functionality, as well as performance. But we have some pretty cool stuff here to actually start pushing the envelope of flexibility and new performance as well. Yeah. Well, let's dive in. Yeah. I'm ready to go. We've got kind of three pillars, because we love working in the rule of three. It helps people tie things together. Obviously, first and foremost, right, the performance, the efficiency side, something that's always been a hallmark of what we do here at Everpure. And I want to bring up our friend, the FlashArray XL 190 with Turbo. It's a really interesting one that serves some interesting use cases, but shows what we can do architecturally with the flexibility and engineering that we have with our array family. So just for those who don't know what the XL 190 is, it's the top end of the FlashArray family. We announced it last year as kind of this ridiculous performance density, right, like you could go, you know, the scale-out multi-engine configuration we can do in just a little five rack unit configuration, which was amazing, right? And it came with a bunch of DRAMs so you can do read caching to give you even lower latency. It was about 50% of the latency. What we're doing here with Turbo is a little bit different. We are changing the paradigm a little bit. So in the past, we had an active and standby controller, and that standby controller, we were actually starting to add some capabilities to it, some background data reduction features and stuff like that. What we're doing with Turbo is we're enabling incremental headroom. It gives people flexibility so that if you're running highly transactional, low latency, let's say Oracle or databases that you're running on that system, the last thing you want is like a backup workload or an analytics workload or even an AI workload who's doing very read, you know, transactional access to impact that production environment. And so what we're doing is giving customers the ability to flex. It's got a little bit of Turbo on it, and it allows them to use the secondary controller for read operations. And so it extends the performance, just giving you more agility and more flexibility. And of course, you can buy more systems, but this is nice that you can flex a little bit even on existing single systems for workloads to ensure they're not impacted. Yeah, it really hits one of those corner cases for just those super, super latency-sensitive types of situations or use cases, and again, shows why and how we're always listening to customers for areas that they need some optimization, which is super, super exciting. Speaking of optimization, our friend FlashBlade Exa has been really wowing some folks out there, and we're seeking some great benchmarks. We're looking at things around spec and ML perf, and we won't hit the numbers directly. We'll invite people to go look at that. But it's really super exciting to see this thing that we conceived and then brought to life and kind of blew people's mind with the sheer scale. But there's also, I think, a halo effect for FlashBlade in general, because we know Exa's not for everyone, everyone out there, like there's a small case, but a great halo effect and just showing how we're really pushing the boundaries of performance. Yeah, I think the numbers that we threw out in spec and ML perf speak for themselves. It was number one in the market. This product was really kind of came from a genesis of we knew we needed a lot of GPUs to be attached to this thing. You know, FlashBlade pretty much covered most of the market, but when we got into GPU clouds, they were talking about tens of thousands and hundreds of thousands of GPUs at tens of terabytes per second of bandwidth. And so we built a beast, and we're seeing it in the performance numbers. For those who don't know, we built a BU here that actually manages the entire thing now. So we are hyper-focused on continually adding new feature functionality, both market-driven as well as customer-driven. And so that product's becoming a pretty killer product for us, especially in AI, and I think you'll see it expand out to other use cases as time goes on. Yeah, particularly as you look at some of those big, giant GPU farms that are getting built out, and they kind of come look at this and go, oh, yeah, I think this is a good fit for what we need to do, right? It makes a lot of sense. We're really at the barriers of extremity here, but I think the other really huge innovation that I love is to our subscription portfolio. And here we're looking at Evergreen One, and we're looking at something that we call OverDrive because let's face it, in this day and age, right? You talked about those pesky AI workloads when you were talking about FlashArray coming in and doing some disruptive things. In this day and age, I think it's really hard for our users to predict when you're going to get a performance spike, when you're going to need to get that boost and do that bursting to get more, and now we're able to offer that capability in a subscription. What in the world? How are we doing that? I think it's really showcasing the value of Evergreen One. I think it's going to – you'll continue to see more stuff like this where it gives the agility of a traditional cloud environment, but on-prem with sovereignty and all the great benefits of the SLAs that are behind it. But just the benefits – we talked about adaptive tiers last year. This is one to just increase the performance when you need the incremental headroom as well. It's a really good time, not only with the supply chain things going on for Evergreen One with getting consistency of pricing and the like, but just another cool feature. I love Evergreen One. I think the SLAs are just so killer for customers to get commitments from us on it, and to have some of these cool capabilities of agility, it just makes it even better. Yeah, and it's a feather in the hat of that risk reduction thing, and particularly it helps you connect and provide some security to the CFO in the organization from a financial standpoint because that is very disruptive if all of a sudden it's like, oh, we're having a performance spike. Oh, we need to go into another procurement cycle, and how long that takes, and maybe it wasn't budgeted for. And so here you can take that spike and do it at a subscription level and just have that built in. Risk reduction. Good idea. All right there for you, which is great. Okay, pillar two, talking about data depth, maybe the term access, right, being able to access the data you need, where you need it, when you need it, and what we spend a lot of time on here is really in and around the protocols and just advancing the capabilities around there. Let's start with the enterprise data cloud and file specifically, right? We've always had this great replication technology, super simple. I came on board and it was just coming out, and I was like, what do you mean only four steps to connect and to do a metro cluster and do all these things? File is now an equal citizen in that, and that is super exciting to add that into the mix because file is complicated. Oh, it's incredibly complicated, and it was very hardware centric. You'd replicate these ports to these ports. What we wanted to do is take that same workload construct, that abstraction we talked about in the last section, and now make that the paradigm of which we replicate between the two. And it takes the infrastructure and again abstracts it away. We also added in the benefits of presets or policies for file. And so you kind of get this full fusion experience with file now. We started with Brock, which of course was kind of like the main use case a lot of customers use it for. File's been growing like amazingly well. And then we added it with Accent Cluster File, which is giving you synchronous replication across two, but it's a workload-oriented configuration versus a hardware centric. Yeah, yeah. And a big difference there. And probably how people want to operate more as well at that workload level or at that intent level. Yep. Coming back to the whole intent thing. It comes back to that. Yeah. This is what I want. Go deliver it. Yeah. Yeah. Yeah. Which I love. Staying on the unstructured side too, there's been a lot of movement forward with object, right? Yeah. It seems like I can't, you know, open up a – I was going to say open up a newspaper, which makes me sound old, but, you know, click onto an IT article and not see something about just what object is doing and straining storage teams because of that proliferation of unstructured data. But we're on top of that. We're seeing object pretty much in every possible use case. One of the bigger growth ones are like modern applications. People are building new apps that are object pretty much entirely, right? Yeah. Like, they're putting their persistent data there. Their index is there. They're querying it directly. Really changing the paradigm of the application infrastructure. Last year, we kind of showed this object vision. Yeah, yeah. We showed FlashArray, NowHud object, FlashBlade also. We're adding in a couple of cool capabilities in object in particular. The first is, you know, tagging, which is one we've been wanting to get done. Being able to do workload – sorry, lifecycle policies so that you can define where things land within the system. We have capabilities of strong consistency we'll be delivering at the end of this year, which allows you to have, you know, both systems be fully consistent across the two, which will be a big win. So it's really just expanding the opportunity for us to be able to take that edge core cloud configuration and help customers be able to deploy applications, again, in this very workload-oriented way, but do so wherever data resides. That's the big win. And doing it as a fusion first, right? Of course. Like, everything fusion first. I love that. That is great. Okay, so that's pillar two. Now, pillar three, we're going to have to wind through some things. They're related, but really about workloads and people's ability to run applications, but the notion that what we hear people want is flexibility. Now, some of this driven by some of the challenges with massive changes in modern virtualization and what's gone in there. And so that's certainly a couple areas that we'll cover, but then also a little bit of peace on AI at the end. So we'll just flick through and maybe hit the Azure local piece first, because I kind of hit that on a podcast with Quimby a little bit back. But it's still fresh. Oh, yeah, it was hard to keep him from going off on Jags for eight minutes on things, because he just knows so much. But it's a super exciting advancement and something that people are starting to look back into a little bit. I think sovereignty is driving a little bit of this, right? And so I think as people want to be able to leverage existing Microsoft services, having local infrastructure or running it at the edge in many cases, having FlashArray as well as FlashBlade actually supported in these ones is a big win for customers, because they can get all the benefits that they've been enjoying with their existing on-prem solutions as well. We're doing a bunch of work around Microsoft in particular. Obviously, EverPure Cloud's been built out, and so we have a bunch of cool stuff that's coming out there for Azure VM native solutions. And so it kind of just continues that overall story of cloud all the way out to edge and back to the core if they're running Pure directly. Yeah, and we had a lot of great sessions at Accelerate since we'll publish this after. So please, folks, go back and, you know, David Stammen and Nick Scuola, and they're just talking about all the great things that we're doing in this space. So check out those replays. Another one I wanted to hit on, and what I love, is when we forecast something on the podcast. Like six months ago, I had Cody and Ketan from Nutanix on, and they just said, look, this is just the beginning. Like, we've got foundational things to get people going, and it's highly functional, and the partnership between the companies culturally is meshing. But we're not stopping here. We're bringing out new things. And one of the goals is to just make that experience that people have been comfortable with with their virtualization environment very similar to what they're going to be doing with AHV, and now we've got new functionality, like vVols, right? But this is all about the Nutanix relationship continuing, and we're six months after we talked about that, and here we are putting out new capabilities. It's exciting. Yeah. I'm excited about the partnership. I think, you know, one of the things Cody and the Inge team, Pavel, and the whole crew has been building is just helping customers get new values of being able to transition between different hypervisors. And I think that the vVols one on Nutanix, as well as some of the stuff we're doing in, you know, Portworx and conversion, we call it Teleport, has been really showcasing to customers that we can build truly innovative technologies inside the box to do conversions fast that would take them hours to actually do migrations that we can do in mere minutes. The benefits of that is we're building, you know, unique technology for them to move faster, to get wherever they're trying to get to, and provide an experience that others just can't deliver on. Yeah. And we're doing it multiple ways, right? So the other part of this that's really exciting is the momentum we continue to see with Portworx, right? And, again, acquisition from a few years ago and very container-centric, but what we've latched onto really here with this market moment that happened a couple years ago is it almost is compelling people to, like, take that jump forward and get into containers at scale because it's that alternative for putting VMs into containers, and you've got a great partnership with Red Hat and with SUSE and using KubeVert. And so we've got some great new innovations coming out in that space as well. Yeah, this is a fun space. I think we're continuing to evolve. The Teleport project actually won. We showed a little bit in New York last year of doing kind of a VM to KubeVert migration. That was such a cool demo. I still talk about it. Yeah, it's a fun one to do all via Copilot, too. As much as I possibly can, I want the entire platform to have that kind of experience, and we're going to get there. But we've seen some customers who have asked us, like, hey, we've got RDMs on a big SQL server. It takes us, like, five hours to actually migrate to KubeVert, and we're showing them it done in, like, five minutes. And they're just shocked by it, right? It's the innovation that we're building in back to the core, you know, features and purity, like, you know, doing, you know, copy mechanisms that allow them to not actually have to do a lot of data mobility and a lot of migrations. And so these types of innovations are fun to showcase to customers because I think they kind of see the deep innovation we like to build with hypervisors, like we did with VMware, but now we're doing it with alternate ones. Yeah, and it's almost like those old choose-your-own-adventure books from the 80s that we all loved, right? And if you're much younger than we are, go check those out. Those were fun, but we're providing that choice because we recognize everybody's in a different situation, different business conditions, different budgets, different resourcing, staffing. And so here's multiple ways, which, by the way, also stay on VMware and we're still innovating on VCF, too. That's another piece of it. So it just shows that multidimensionality that we have both as a partner with those alliances, but also the scale that our engineering can do to go down different vectors, but also ultimately give customers choice. And one other great thing is like Portworx provides an experience that's somewhat very different than anyone else can deliver in the storage realm of things, but tying it back to all the stuff we've been talking about in all these other podcasts of it being enabled with Fusion and its understanding of policy and intent. And a lot of the other things we talked about, though, the mobility concepts and the like are all still very Fusion-oriented. And so customers aren't having to move away from what they have come to love in this new version of Pure, but they get all these inherent benefits that Portworx brings to the storage management side of things. Yeah. It's like good improv. Yes, and. Yes, exactly. Right? It's all great yes, and stuff. Super exciting. The last one I want to close on, and because this is going to tie into Episode 3, right, where we talk about EverPure Data Intelligence, because there is that lifecycle of how we're treating data contextually, but the unified data plane now has the ability with Datastream to treat and prepare AI data. And I think I've referenced this one on the pod before, you know, with Alvarez, but just to hit that is now something that's out. We had a session at Accelerate that Casey Lai did, super great, great work by the team. But this is another piece of the puzzle that allows people to have that workload flexibility and that unified experience, but the platform. Yeah, I think a lot of things we're going to continue to talk about is just getting data ready for AI, understanding the semantics of your data so that agents can rationalize it. I do think the future holds, and Ashish is going to talk a lot about this, is like this new semantics layer becomes the new substrate for agents and analytics and a bunch of other things to actually consume the meaning and understanding of data. But if you think about just the existing model of your current data, it needs to be also understood and vectorized and, you know, available for RAG so that your queries on existing LLMs come back better with your actual data versus its own memory. And so that's where, you know, Datastream really plays incredibly well. It allows the data to sit in place. You can vectorize it and make it available. You have a vector database for models to actually be able to use for RAG, and it's a killer solution. Yeah, it absolutely is, and a good bookend to all the different things. I can't believe we waded through all those here. It was quite a bit of things. In the time that we had. Yeah, I feel like I'm going to pat ourselves on the back and say I think we connected them together pretty well. Yes. It's super exciting. So, to that note about bookending, just give people kind of the summary on Unified Data Plane and why you're so excited for that pillar of the company, right? And we really were very deliberate two years ago about shifting to talking more about platform. And I think the Unified Data Plane really is that manifestation of our feeling about what we're offering to customers and how we continue to innovate for that seamless experience regardless of what the data is or where it is. I think, you know, if you look at our heritage of building a platform that not only provided everything non-disruptive, but was evergreen in nature to swap out any components, this was this natural evolution to build something that acted like one holistic system, but did so via specific core competencies, right? Like, we talked about a bunch of features here of agility where you could add more performance, overdrive and turbo. We had capabilities of being able to think about things differently, even in traditional models. Synchronous replication has been around for a long time. But do it in a world of workloads and abstract away the core infrastructure stack. Think about data holistically, you know, cross all of these, you know, edge core and cloud, but do so in a manner where you can rationalize all of it as one holistic thing. And then as you start thinking about the flexibility of these workloads, being able to take on new things, not just databases, but be able to actually understand and optimize things like containers or conversion from one hypervisor to another or vectorizing data. These capabilities are just amazing to think about it because when you think about it, you have one common data plane that can do all these great things. And then you layer on top of it, this amazing intelligence to govern all of it. And then we're layering on top of that, a semantics layer that takes the meaning of not just pure data, but all data, right? And that I think starts to really change the game for the way customers think about how they think about their data, as well as how they think about Everpure. You are welcome back on this podcast at any time with that excellent summary. No, you put it together. That was exactly what I was looking for, to get people out the door and why we have you keep coming back. Awesome. Because that really encapsulated it and everything that we hit. Thank you again. It's always a pleasure to get you in here and we always have fun. For sure. As I said on episode one, so it was super cool to do these. Have a great summer. You got any vacation plans in the mix? Not a whole lot. Not a whole lot? We just got back from Europe. Yeah. That was a big trip for us. Yeah. We'll probably do something in the August time frame. The kids have camps and all sorts of stuff. Sure. Sure. No, you've already pre-scheduled those things. But Accelerate's the big, you know, runoff we're going for here. So excited to see all of you out at Accelerate. Hopefully many of you join us. And I love doing these. Yeah. Thanks for having me on. Thanks, man. Super great. My big thanks to Chad Kenney, our VP of Product Management, for coming in, doing episode one of launch for the Intelligent Control Plane, what we just walked through with Unified Data Plane. And if you liked these, I would invite you back for episode three that we'll be publishing shortly with Ashish Gupta from the company formerly OneTouch that we acquired, and our ever-pure data intelligence. And I'm super looking forward. That's going to be a big challenge, too, because that's an area I've got to do a little bit of learning. I learn a lot every time I talk to you. I know. I know. I was just talking to him the other day, and I totally learned a lot. So thank you out there. Thanks to Chad for coming on. And thanks. I hope you enjoy these series, and they just pull together everything that we've been doing because I know it's a lot, but it's really important that we let you have an understanding and a flavor for everything you're doing. So please, as always, share out this episode, tell a friend, share with a colleague, and we will keep the great guests like Chad coming on to the program. With that, we will wrap for ever-pure for my good buddy, Chad Kenney. This is Rob Ludeman saying, don't look back. Something might be gaining on you. I know. I know. I know.
