00:10
Good morning, Pure fans. Thank you so much for coming and joining us, uh, these, uh, these few days in Las Vegas. Um, is it hot enough out here for you? On our way in 110 degrees, uh, you know, the smoking, right?
00:26
And that's what we hope, uh, that we're doing here in Las Vegas is we're really smoking with all the with everything that we're doing in data storage. So first of all, I like Lynn, I wanna thank, uh, everyone for coming in, all of our customers, our prospects, our partners, uh, our, uh, press and analysts that are here with us today,
00:46
too many to mention, but we. We really are very honored by you coming out all this way, taking your time to spend with us, and we hope that you come away from this feeling more enthusiastic than ever. Certainly we have. It's because of you that we pure have grown consistently every single year, the fastest grower and the in some ways the only grower in enterprise data storage.
01:12
Consistent growth every year, the only. Western vendor to grow every year and that's what's catapulted us now to $3.2 billion in total sales, uh, very profitable, 17% operating profit. Uh, another statistic that might be surprising to some, we're almost 50% now, uh, of revenue in our subscriptions.
01:34
This is Evergreen forever, uh, as well as Evergreen One and an increasing percentage of our sales as, uh, going as a service with Evergreen One. And then the number that I'm most proud of and that really sets us apart from every other vendor is that we will spend even this year more than 20% of our revenue on R&D and if measured in a, on a um gap basis over 25%.
02:00
And it differentiates us because it's pure that has always considered data storage to be high technology and not a commodity, and that's what's allowed us to stay number one in innovation in data storage and you'll see on this Gartner Magic Quadrant yet again this year, number one, what was different though in this year is this is the first magic quadrant. That measured data storage platforms and they put us again as number one in in data
02:31
storage platforms and it's because of our very consistent, very advanced and reliable data storage system uh based on purity. So you might ask, OK, we spend 20% or more on R&D. What has that gotten us? What has that gotten gotten you? And I would say that we've reinvented storage in many different areas.
02:53
We've reinvented storage, you know, for the cloud, we've reinvented storage we believe for AI, and we're again reinventing storage for enterprise, uh, especially in this new era of data. All right, so let's talk about um AI now. Purity, again, is the basis of what we're doing in the AI space.
03:15
We've just introduced and we are now, as of this show going general availability on our newest flash blade exa product. Flash Blade exa is just another variation, if you will, hardware variation of our flash blade product. Flash blade, of course, very, very fast file and object storage,
03:35
where we provide both the protocol support as well as the storage inside flash blade. Well, with an update to purity. Now flash blade can operate disassociated with separate data nodes, and by operating with separate data nodes, we can increase the overall performance of the flash blade product.
03:57
As a matter of fact, Flash blade as well as Flash array is the fastest metadata uses key value store to create the fastest metadata engine in the world. And by extending it with additional data nodes, we can allow parallel performance for access for high performance data such as for AI. And then by adding additional metadata nodes and additional
04:24
data nodes to that. We can consistently increase in a linear fashion, the overall performance of flash blade exit. And this linear performance we've proven now over the last several months, and it just keeps getting more and more performant as we add metadata engines and data nodes.
04:46
This is completely what we, uh, as we introduce it today, it'll be 5 times faster than anything else in the world. At high performance AI for large scale systems. And because it's based on purity, we're now able to scale all the way from Flash Blade S, which many of you are using today.
05:08
That can already handle many tens if not hundreds of GPUs in a cluster for AI performance, but for those of you that may worry about scaling to thousands of GPUs, if not tens of thousands, flash blade exit can take you all the way up to 100,000 GPUs. And then with meta meta with a different type of architecture uses we'll be using our direct flash technologies as I spoke about before for even larger scale environments.
05:39
So again, the promise of purity of purity plus evergreen, we provide you again the ability to scale not disruptively from the smallest workloads to the largest, from low uh from lower price all the way up to the highest performance. Now, let's talk about AI though in, in the real world, in the enterprise world, where AI, you don't worry about necessarily going to 100,000 GPUs.
06:06
How many of you are gonna be going to 100,000 GPUs in the next year? I thought so it was the only question I, I, I've ever asked on stage where I didn't expect to see any hands. So AI in the enterprise space we think is a, well, we all believe that AI is fundamental change. It's changing our jobs.
06:25
It's changing the jobs of our colleagues. It's going to affect really every aspect of IT job categories, workplaces, workflows, knowledge workers are going to be disrupted. But you know, one of the things that I think is interesting to think about is that AI is gonna change the relationship between software and data. If we think about the last decade or two, it's really been software that's been eating the
06:50
world, right? That's, that's what Mark Andreessen said. Software is eating the world. Data was necessary, but software was the key. And let's just think about ride sharing, right? Is that that's been a big disrupter. We used to take taxis or, you know, uh, black cars when we would go from one place to another.
07:08
Now everybody does ride sharing, right? Except If you go to San Francisco today. Do you know that more rides now are taken in self-driving taxis than by rideshare apps? It's pretty amazing, right? Well, what's a self-driving, uh, taxi? What's a,
07:28
what's a self-driving car? It's all based on data. It's been trained on data and the quality of data is is what's made the difference, and that's the difference that AI is making is that the data is becoming in some ways more critical than the software itself and that's how so as we as workers that work with data look at our lives going forward,
07:52
look at what's important for us going forward, we're gonna be data centric. So let's talk about how data is kept in the enterprise. Let's look at enterprise data and data and enterprise data center architecture. So if we go into a data center and we start building a new app, or we start accommodating uh a uh a new type of app, let's,
08:15
let's look at, for example, a new database environment, right? We start with compute, the software that goes on to a um uh a compute node. And then of course, you're going to add um storage to it and a network to connect it to everything else. And this is a stack, right? Now, The server may be virtualized and and you may have a VM or more modern maybe going into
08:41
containers, but typically the storage is direct connected to the application stack. It could be, of course, more than one server, but it's dedicated, it's sized, it's managed. Uh, if it's fiber channel, it's it's literally directly connected uh to the application itself. And then you build another application or you accommodate another application,
09:03
um, and that has its stack. And another application or backup and it has its stack. And each of these environments is separate from one another. And if you look at that and look at the way that. The data storage is managed.
09:20
Well, the arrays are managed individually, right? And the um the set up, the provisioning, um, is, uh, 11 by one. you cannot share the capacity between two arrays. You can have two database apps right next to one another with different arrays. One array can be full. One array is, you know,
09:40
only partially full, but you can't really share the capacity between them. They've been designed, uh, you know, specifically for their application environment. Um, your, your. Your servers may be virtualized, but your uh storage is not right and governance is array by array and highly manual, you know,
10:03
so that is your snapshot policies, your backup policies, your resiliency policies. You are configuring that array by array. Now, that's looking at it from a storage standpoint. Let's look at it from a data standpoint. The data itself is captive to the application stack.
10:22
Two ways to get access to it, you can copy it, or you can go in through a set of uh connectors through the application. You have poor records of data copies and movement. I know this for a fact because we've seen it everywhere. Anybody can copy the data, go somewhere, no records, uh, kept of it, right?
10:43
or any records are manual. The raw data itself is inaccessible to other applications. Again, you have to copy it to get access to it and data governance is inconsistently applied. To these environments because it's applied manually by people, by people in different areas and exactly what has been put into
11:06
place is rarely standardized. If it's standardized it's standardized on paper but not in software. So let's compare this to the way clouds operate, right? What is the architecture in a typical hyperscalar? Very different. First of all, it starts with,
11:27
well, we could start in multiple places, but let's start with the compute layer. They have a layer of virtualized compute, right, that can be assigned, you know, in software, right? But similarly they have a layer of storage. That layer of storage is the same across all of the software environments.
11:50
It's also virtualized now they may have multiple uh different layers of storage. Those layers are really uh qualified by price performance. They'll have a high performance tier, they'll have an archive tier, and they may have a couple of tiers in between. That represent price performance layers, but outside of that it all looks the same and it's
12:12
shared across all of their application environments uh all across all their customer application environments, and you get access because it is coordinated through a, uh, management layer that assigns how different storage or how different compute is assigned to different, um, application workloads, right? And think of it as instead of full stacks, which are sort of hardware.
12:37
They are virtual full stacks, right, which is all based on software. So let's just compare these two for a minute, you know, the enterprise architecture and the cloud architecture. Enterprise storage architecture is vertical. Vertical stacks. Whereas cloud, uh, storage architecture is
12:57
horizontal, right? It's a, it's a common layer that is assigned in software. Enterprise storage architecture is highly manual, cloud storage architecture is fully automated. Cloud, uh sorry, enterprise storage architecture creates data silos. The data is captive to the application stack, whereas cloud storage architecture creates
13:21
accessible data pools because assuming the appropriate role based access controls and security, it is physically accessible to other applications, which makes it easy, easier to get access for things such as uh AI. Um, so I think of of enterprise storage architecture as being physical, whereas I think of cloud storage architecture as being virtual.
13:49
And so this thinking has inspired a pure. To say, well, how can we make the enterprise data storage architecture that exists today. To look more like a cloud architecture. And this is our vision for what we're calling this new era of data. It is a new era of data where data is becoming more dominant.
14:13
Relative to everything else, our jobs of managing data is gonna be critical to the success of our organizations and so as we look at that, we think of it as um requiring a much more consistent environment, a much more automated environment for data, and we think that in order to do that, it starts with a consistent software layer which we call purity.
14:37
Now you know purity, it provides uh block file and object. Uh, it does so at every different price performance level, and we're able to do that because Purity at its heart is just a giant key value store. That is able to handle any type of of data as a key value source, so it's very fast block file and object are all handled in a very similar way and of course we
15:04
have built in all of these ilities if you will, you know, whether it's resiliency, um, it is upgradeability, uh, it is um uh uh the evergreen capabilities we built that all into pure and we've put it into a platform that is. Really quite extensible so we think of this, you know,
15:23
it's common operating properties. It's the evergreen capability and all of these uh capabilities are built into it and the fact that we can now scale from the lowest price, you know, from even getting to the same TCO as hard disk environments, um, all the way up to uh the highest performance in AI. As well as with all of these ilities that are built into basic purity,
15:50
this is what we mean by a unified data plane. So starting in order to better manage your data in an enterprise environment, you need to start with a unified data plan. But the key message that we're bringing to you today is we've added something that is very new and very different.
16:09
All of this is, uh, sorry, I got ahead of myself. All of this is managed by pure one, all of it available as a service, and all of this able. To manage all all of your different uh workloads whether those are your enterprise apps, the traditional apps of VM based apps, or whether these are new apps that are based on
16:28
containers, AI and ML, uh, or, um, you know, the latest in terms of Kubernetti's virtualization, uh, which we provide, uh, both with Purity as well as with Port works. Now the big reveal, sorry about that is the new thing that we've added is this new capability we call fusion and what fusion does is create a new intelligent control plane. So for those of you that have ever been in a network, you know what a control plane does.
16:57
It allows individual, um, in this, in the case of communications, routers and switches to operate as a network. What we've done is we've allowed all of our individual arrays to operate as a cloud of data, an enterprise cloud of data. It allows all of your arrays to be managed as a fleet.
17:16
But more than that, it allows you to define how you want your different data categories, your different data sets to be managed on a global basis. This is not about clusters. It's not about just allowing a few arrays to operate together. This is about a global data cloud with different availability zones availability zones in regions that you get to define.
17:42
And that will then apply your uh your global governance and um uh compliance standards on a global basis and operate on that data automatically. You define the policies that all gets set automatically. So, let's just look at those two things. The difference between individual arrays and an enterprise data cloud.
18:09
And uh global uh individual arrays are manual. They require manual provisioning, whereas a data cloud is auto provisioning. You go from a defined capacity and performance on a per array basis and you go to shared capacity with uh with uh with auto load balancing across them so saves saves you money. You go from manual governance to software based global cloud governance of different data sets
18:38
that you define. You go from the uh protection that you have to set up manually array by array. To protection that is then done by the policies that you set and you go from uh data that is captive to the arrays to data that can now with the appropriate governance and authorization to be uh basically acceptable uh accessible by any application,
19:05
and this is what we mean by you being able to build your enterprise data cloud. It means going to a virtual cloud of storage rather than dedicated storage per um uh per array. It means going to a consistent environment across your global network. I know how that operates.
19:27
We all know that we are not consistent on a global basis, maybe not even within a single data center environment. You go to automated data governance. Automated, think of that you set up the governance and if you ever change your compliance standards or your governance standards instead of having to go to every
19:47
single array and make changes, you just change the uh the presets and it happens automatically. You go to built in cyber resilience because it just becomes part of the preset rather than doing it again array by array and then finally you become software defined, you become software governed, uh, and you provide a global scalability of your governance standards.
20:15
This is a major change not just in the way you manage storage, more importantly. It becomes a major change in the way you handle your data. So the way to think about this is when you build your enterprise data cloud, you can start to stop managing your storage and you can start managing your data on a global basis.
20:41
So we have 3 speakers that are going to take you through every element of this. We have Chad Kenny that's going to be taking you through this concept of an intelligent control plane and what it does for you at many different layers. Rob Lee will give you all of the updates on that unified data plane because if we don't support all of your different use cases and workloads, of course it doesn't really become a
21:04
full enterprise data cloud. And then we're going to have Naveen show you how this can be done as a service for you, so that in fact we do most of the work and not you. So I wanna thank you for your time and I hope you enjoy and I'm sure you'll enjoy the rest of the show. Thank you.
