00:04
Thank you all so much for coming for those of you coming in the door right now. Hey, you made it just in time. That's awesome. We are here to talk about like it says going deeper on pure accelerate keynote announcements, fleet management automation A I driven resilience in case you didn't figure out already, this is not gonna be quite like some of the other sessions which are wonderful.
00:25
This is gonna be more discussion and wandering through some of the topics up there with folks that have spent the last 23 years of your life working on this for Larry with Fusion Tago. You've wandered around pure, I'm not gonna introduce them. So this is gonna be more discussion based. We're gonna wander through some various themes. I've got some notes and, and for those of you who don't know,
00:43
I'm gonna move into a little bit of introduction. So my name is Andrew Miller, lead, principal technologist here at Pure, I work with customers, partners, uh support customers, although not giving articles of clothing as we started out and established already. So, so sorry. Um But in this case, we, we uh I,
00:57
I've been accelerated for a couple of years and we, what we're gonna try and do is wander through in a little bit more of a discussion format, some deeper perspective into why and how we created some of these things. What are some of the benefits for you? I also host the monthly coffee break series. If you, some of you have joined that, I appreciate you joining.
01:13
We usually have one or 2000 folks join and we w through some of the same themes. I actually see uh John Brett out here, I had him on that in the past. So before we continue into what we're gonna cover today, I actually want if you don't mind, Larry and then Taga, do you mind introducing yourself and what you do?
01:27
Pure? Yeah, sure. My name is Larry Touche. Uh I joined pure almost 10 years ago. So I've been here quite a while. I joined as a technical marketing type of uh person working on async replication. And then I became the, the product manager that launched uh active cluster.
01:44
I worked on the Flasher AM to uh XNDU series and the original FA 400 to M NDU that we did as well. Um I like to tell people if there's things you like in flash array, I might be responsible for some of those. If there's things you don't like, I might also be responsible for some you didn't cop out on that. Sometimes you have to make tough calls and they
02:07
don't make everyone happy. But that, that, that's me. Uh these days, I'm the product manager for uh fusion and my name is Santiago Nava and I go by Tago. Uh that's not written up there. It is Tago. Um I've been up here for a little bit less long and uh then Larry,
02:23
not quite 10 years, I think it's, it's a little over five. I used to be responsible for some of our hardware products. So you may have seen me in a flash blade platform, uh role flash blade hardware. We launched flash blade s uh two years ago now. Uh and the E product family last year, but most recently over the last year,
02:40
um I focused on our evergreen one product portfolio. So today, I'm responsible for uh that part of the offering and really the storage as a service uh offering that is awesome. And I think we're pretty sure while we're gonna wander around topics, Larry knows stuff about what Tago does. Tago knows stuff about what Larry does all
02:57
common in general. This is gonna be a little bit loose and relaxed and I'll try and make sure that I actually look at this part of the room too and you know, we'll address everybody. So to kick it off in a little bit, you saw this on stage earlier today. So there is a good bit of um focus and messaging and we're actually delivering things
03:13
around a platform message, right? So we have things that there's the core Evergreen architecture, unified Infrastructure, a common purity os with different ways that it's surfaced and flash and flash blade management with simplicity has to be a continuing focus. We're not gonna go as deep on this, to be honest as there's also the piece of when we
03:33
look at this and now you're like, oh, I'm gonna stop listening. I'm just gonna read what's up there. So you actually see this in some of the launch blogs. But there are a bunch of things that we address. What we want to do today is explore it with some depth underneath the pillar. So there's an overall platform message and you'll see a pillar around innovations in the
03:49
pure storage platform A I success in the enterprise storage as a service. So we're gonna kind of wander around this a little bit. When we talk about fusion, we might have a visual for Fusion. We talk about Evergreen one or as a service. We might have a visual for that. We'll kind of leave this up here a little bit as the framework.
04:05
If you're hoping for lots of slides, sorry, there's other sessions, there'll be a little bit more discussion than that. So that may be I think when I think this was actually Utah and when we were talking ahead of time about kind of the common themes around both between what you and Larry do around simplifying operations, getting the faster outcomes.
04:25
Do you mind kind of starting off there and then take us into some of the overarching? Thank you. Yeah. And I think that really ties a bit to the uh platform message we're trying to give, which brings both of those things together, which is therefore a good starting point for us. But when I look at the evergreen one storage as a service portfolio, when I look at uh some of what Larry is driving with pure fusion,
04:46
both of them have that common thread of trying to simplify operations and trying to focus on uh delivering on, on certain values and certain outcomes on a slightly higher level, right, or slightly more complete, more holistic level. So I think pure Fusion and Larry is gonna go into detail on that maybe but pure fusion does that much uh you know, further down the stack at a technology level,
05:09
really allows us to unify the the core components of that technology into a unified platform while Evergreen One attach uh attacks that from the other side, right? From the top down, looking at how can we turn the great pieces of technology that we have, which fusion is one into more um outcome oriented uh topics for a customer? Right.
05:35
So how can we deliver on what you're looking for leverage on our technology? Both of them help you manage storage or think about storage without having to think about an individual array, either a management layer or consumption layer. Anything else you want to add in there, Blair before we kind of dig into fusion a little bit. Uh No, I'll dig into the next boom. Let's do it.
05:56
OK. So you said the last 2.5, we've been talking about different pivots and even, even some fun, fun behind the scenes, stories and periodic meetings with Charlie because he cares about fusion. So those are, those are never not stressful, but you've been very visible on this. So when you think about automation and management at scale,
06:15
that's one of the key themes of fusion unpack that for a little while. So um there's a lot of uh so I just did my um fusion session about an hour ago and I see scheduling, it would have been nice to direct some of you into that session. Um The scheduling goes backwards. But yeah, I see a lot of faces in here that I didn't see in, in that one. So I'll give a brief kind of overview just to
06:38
make sure. Yeah, yeah. So um uh so you may, so you may have heard last year we shipped a version of a thing called fusion that was hosted in the cloud. It's a cloud based control plane that connects to uh arrays on premises and supported Izzy and it had workflows in it to simplify storage, provisioning. Um But it didn't yet support replication and
07:02
things like that and it required that your arrays were connected to the cloud and, and had you had that communication channel from the cloud management and point down into the arrays. Uh the fusion that we're shipping this year is the next evolution of what we're doing with fusion, which is taking it directly into purity. So starting with the on the flash array side, starting with the 6.8 release line fusion and
07:27
its capabilities are just part of purity. There's no separate license or subscription. You will see a new uh new features and functionality show up in the cli the gooey and the rest API on the flash array when you get to that version of purity and you're free to turn it on and use it with no cost or no license. Um It'll come to flash blade uh as well early next year.
07:51
Uh I think the target version for that is 4.6 something. Um And what fusion is really trying to do similar to what Tago was saying is automate things for you. So one of the key features that'll come in fusion uh near the end of this calendar year is the thing we call a storage preset. And the storage preset is a way for the storage administration team to declarative define the
08:16
way they want storage to be configured like for an oracle database or for a sequel database or for a VM ware data store. Um or for any kind of workload, you can think of a cross file block, an object you can define the way that storage should be defined including Q OS properties, data protection, replication, snapshot retention,
08:38
custom labels like billing I DS for chargeback show back define all of this in the preset and then you provision from the preset and fusion will work to do all that set up for you in the background. So if you saw um Sean Hansen's keynote message this morning, he showed you the little business card. Um um The I'm with stupid one or the other one, the other one.
09:01
Did anyone else catch that one? Yeah. Um The uh um So over the years, our business card has changed from one card with two sides and then we added a few more features and it became a, a foldable card. Um And then uh when we shipped active cluster, I was actually the person who was tasked with
09:18
making that fit on um three cards. So we had a tr fold. Um The one of the, you know, themes of what we're going for now is that if you think of managing your environment by defining the outcomes that you want in these simple storage presets, um We can shrink that business card back down to just two sides, right? Just set up the the outcomes that you want and
09:41
then provision from that thing that gives you those outcomes. There's two things I think I want to highlight there. So a customer for a while, partner for a while and sometimes I'd attend conferences like this or keynotes. And I hear about all these cool capabilities and I'd be like, OK, when is it gonna come out? So I said later,
09:54
later part of this year and then will I actually ever be able to afford it or use it? And what will I have to add to my environment? So this is highlighting what you mentioned, no additional license cost. This is built into purity. These are capabilities that you can use, whether you have one array or 100 arrays and you may even want to use it either way because you may want fleet level management or you may
10:16
say I want to do policy driven pres I wanna do presets to use the exact term. I might want to do that with one or two arrays. We're trying to do this in a way that actually can apply across our entire customer base from a technology adoption standpoint, licensing capabilities, et cetera. I think you'd also men if we, I almost wanna just frame it if you don't mind,
10:35
Larry. So if we just went explicitly to folks that are sitting out here, either storage admins, engineers, architects, what are the practical benefits that they would get either right away and then even over time, we can go a little vision too, I think, yeah, I'll start with Um um So this year you're gonna see fusion ship features in, in like kind of two major buckets.
10:57
So the first thing that'll happen in the first release that fusion shows up in is um what we call basic array federation and remote storage management. So it gives you the ability to securely link and connect arrays together if you already have pure arrays and you're doing replication today, you know, you have to type a command to connect two arrays so they can replicate.
11:19
Um We're making fusion so that you can on board your existing arrays into a fleet with workloads in place. And so you will have to go to each array and join them into a fleet. Um We're not just gonna automatically assume what you want your fleet layout to be and change your security posture. So it is an opt in thing that you have to do if
11:40
you want to take advantage of it. Um And that creates that secure communication channel between arrays. And now we can use that communication channel to forward commands from array to array. So in effect fusion sort of like um it like it transforms all these disparate disparate arrays into one sort of platform if you will and you can approach that platform on any
12:03
management endpoint which every array becomes a management end point for the whole fleet. Some of the very practical advantages that come in that first set of functionality is um let's say you're the storage admin, you get a call from a DB A and he says, hey, I need a volume resized on my, on my Linux oracle host. And you go OK, what's the volume name?
12:24
And he's like, I don't know, I don't see the volume name. I can give you a serial number and you're like, OK, I have a lun serial number but I have 100 arrays. Where, where is this? L uh once you've created that fleet, you can log into any array in the fleet. Go to the details page, put in the serial number and fusion will go find the volume for
12:43
you. So it doesn't matter which array you logged into. It'll say, hey, I found the volume over on array number 47 you can click on it, change the size of it, put it in a P group, take it out of a P group, connect it, disconnect it from a host. You can do all the management from whatever
12:56
array you happen to log into and you'll see that in the gooey through this simple little toggle, we call a fleet view. So you land in the Gooey page. It looks like it normally does today. You click the fleet view button and it toggles that page to say, well, I'll show if you're on the volumes page, all the volumes in the fleet.
13:12
If you're on the host page, all the hosts in the fleet and we'll do that for pods p groups. Um All the objects so you can find them and manage them uh from anywhere. So that's like the first sort of practical and all that would work on prem if you're connected to pure one. And is anyone here in a dark side environment or not connected?
13:31
I see some hands. All of what you said applies there too. Yeah. So not even phone home. Yeah. Yeah. So everything I just said, you still get, you can still do everything I just said. Um the next set of functionality um is the workload presets and the preset? It sounds simple when you describe it, it sounds like you're just making a template for
13:52
how storage should be configured. But what that preset actually does is it puts into the system, uh it puts into the fleet a what we call a workload specification that defines what this workload should look like. And then fusion will grab that and go do the work for you to provision the storage and set up all the configuration for you.
14:13
Um And then fusion will actually monitor that workload spec. And if someone says I need another volume on this database, you can provision another volume with a very simple command that just says add one more log volume to this workload in fusion will go look at the workload spec and say log volumes are supposed to have safe mode and this kind of snapshot retention and it'll just do all that
14:36
for you and make it look like the rest of the log volumes for that workload. Um And so it's more powerful than just a uh a template. So I like to describe it sometimes as automation. You don't have to automate. Um But you could still do it all with API S if you wanted because uh you could do it in the
14:55
goo the cli the gooey or you can automate these uh these steps with, with I touch it or scanning the work. If it's um we're not, we're not doing, we're not yet doing the thing where we tell the host to re scan and discover the volume. But we will,
15:16
yeah, we will put the volume uh with the other volumes for that workload. We will give it the proper data protection properties and um we can uh do the connect operation to connect it to the host, but you would, you would have to issue the res scan to make the host find it. So I think, I think we're not allowed to have a session without talking about A I and tying everything to A I and yes,
15:40
that's a cheap joke. But everybody laughed because hey, it's where we are right now. So Tago, you've had some A I injected into your veins in your last role. I think Prakash makes you like do an A I chant every morning to start off the day. Something like that. I'm not sure that's a myth.
15:55
That's a myth. OK. So if, but if you looked at fusion through an A I lens and how it applies, where would you go? And I think so a couple of areas that we should talk about and really, we should tie a little bit of what the, the simplicity that Larry just described, I think on the fusion and going back to that technology layer to what Evergreen one is
16:17
trying to, to achieve. And I and I will give you a spin, an evergreen one spin on this because that is, that is what I do take the challenge. Um And we'll come back to fusion and Larry too. We, we're about some back and forth. Yeah, I think that the connection for in terms of what some of these A I workloads are are looking for. And when you look at Evergreen 14 A I,
16:35
what got announced today at the keynote, what is special about it is not, it's not so much about having another platform underneath. You're familiar with it being the same pure platform that you're used to being made up by things like fusion, like uh flash play, like flash array and so on. Um But it's all about identifying what are the challenges that we're seeing in some of these A
16:57
I workloads and what do we need to do to go solve them? The biggest one that showed up and it's really a common thread between something like fusion and something like evergreen 14 A I. It's finding this challenge around uh the amount of flexibility that some of these workloads are requiring us to, to provide and the amount of change that we see throughout the
17:19
life cycle, right? So if you look at a typical A I typical no such thing I should, I should start by saying, but some of the, the most common A I workloads and uh piece of A I infrastructure that folks are deploying, they do have very well defined and different uh access patterns, requirements and characteristics across that life cycle as a customer uh like yourselves may
17:41
move from, you know, just playing around with the, with uh with A I to maybe doing some mild training rag inference, everything else that comes through it. Um The most common parameter really ends up being that requirement to change your infrastructure as you move from one step to the next, be able to guarantee very high levels of performance in that case specifically.
18:04
Uh but aggregate things move data uh in, in tune consumption in the case of evergreen one over the course of time, right? So I go back to your to your uh question, Andrew around. How does this, how does this tie to A I? And, and to me A I is a good example. It's not that everybody in this room is deploying A I clusters today. But I think it's a good example that we look to
18:29
in terms of what are we able to deliver with something like evergreen one or with something like fusion? And why are these values uh characteristics, valuable uh outcomes for most of you? Even though you're not doing A I is because those same requirements, those same characteristics that we see in the A I world today which everybody including ourselves is making noise about um are characteristics that
18:54
then are hugely applicable to every other workload out there, right? Even uh you know, we we go back to just, just VM Ware uh uh V MS and so on the flexibility, the uh uh ability to change our environment to do uh you know, additional operations on an existing deployment in a way that's easy and that out uh auto tunes itself in a way or looking at the other side,
19:19
not having to change your whole infrastructure just to address another requirement that may have come a little bit unexpectedly. That's what we're trying to deliver um with fusion with Evergreen one. and, and the evergreen 14 A I use case is just a, I think a great poster child for that. I think I may use that if it's all right, Larry, I'm gonna keep going with Tago for a minute.
19:41
Um So we have a lot of things on the slide here. Everyone's gonna stop and they're gonna read that. But, but this is what you wanted it up. So I know, I know I actually built this, look at that. So there's, I'm not a visual designer if I don't know if you can tell we decided to be a little bit, this is meant to be a little bit more relaxed.
19:57
So I see the word enhanced and new and expanded and new. Maybe let's embrace that. What, what's new in evergreen one? You've already mentioned an A I focused sl A but there's more than that. So maybe give us the broad brush of what's new with Evergreen one. Yeah. And I think in the in the spirit of this
20:16
session, which is going deeper on what you've heard around uh ever going one of the keynote, the goal really is to give you more detail on what exactly is new and what exactly has changed and how does this impact or not impact, you know, what you're able to do with the product, right? So that's what, what the little uh new and enhanced and everything else uh uh call outs are trying to
20:36
do or trying to give you a map of how does everything in the keynote translate portfolio? Yes. So the things around our portfolio, we talk about the number of slas and guarantees that we have as part of evergreen one and evergreen one for those of you that are may be less familiar with it are uh our conversation in Evergreen one is really about outcomes. It's about things that you may subscribe to and
20:57
want to guarantee when you subscribe to a storage service, you're not trying to get into um a process in which you are responsible for upgrading your controllers. Whenever you feel like you need more performance. The subscription really means that we should be on the hook for delivering enough performance for what you subscribe to and guaranteeing that
21:17
that is in place at the right time. And by the way, if we mess up it's on us and we pay you right. That's what a service is, is a whole higher level of guarantees and slas that we as a vendor are responsible for such that you don't have to plan um and, and operate operationalize that over the course of time. Same is true for capacity, right? Instead of trying to figure out how much
21:40
capacity you're gonna consume and by the way, how much headroom should I leave for a burst? Uh Instead of trying to figure that out, why not just agree that the vendor should be on the hook for making sure that you have enough capacity in place to consume what you're trying to do, right? That's, that's the paradigm shift. That is key to evergreen one. Instead of figuring it out yourself,
22:01
force us to commit a guarantee so that such that you, we do that for you. And that's why during the keynote in this beautiful slide, you'll all agree. Gorgeous version version. Um It's all about highlighting what are the things that we contractually will commit to you
22:20
as a customer as part of Evergreen one? What are all the things that you'll have a term in the contract says if pure doesn't deliver to this level of service, they owe me compensation, right? They go from performance to capacity, we just said, but some of them are also new. It's the ability to take back a commitment. That's what side we balance is. It's this idea that even though you're
22:40
reserving capacity to get honestly a better rate, because you're very, very confident about where that consumption is gonna take place, things change, right? And you may wanna reallocate that consumption to a different place in which you actually have value for it instead of letting a contract expire and, and pay it for the duration, right? So the side rebalance guarantee is about giving
23:03
you the ability to do that. Uh the cyber recovery resilience slas these are really big for those of you keeping track that used to be known as the ransomware recovery sl A that we launched last year, but we enhanced it. We added um coverage for disasters. In addition to ransomware, we added ongoing resilience services in addition to the recovery
23:26
guarantee, and we've lowered the minimum commitment that uh a license has to have in order to be able to purchase this add on all of that is around giving you a set of guarantees. And by the way, we're happy to continue adding more to this, a set of guarantees to give you peace of mind that if you get hit by something you don't expect somebody has your back, somebody has a bunch of infrastructure ready to go to help you
23:49
recover. You don't have to plan it. You just know that we do when we commit to make that available to you within a set number of hours. Do you mind if I had 11 piece? So we've seen multiple customers. This is the origin of the ransomware sl A where when you have AAA ransomware attack, let's say that you can recover your data back good on you,
24:08
but you don't have anywhere to recover it too because there's virtual police tape around your primary systems. This has happened folks. So even if you haven't had that happen, you should know what would happen in that scenario. Go ask your legal team, go ask your cybersecurity insurance provider. Like what would I need to do in that scenario and at least plan for it and then talk to us
24:27
about how we can help you with that scenario to have a raise on site within 24 hours, 24 hours, right? 24 hour shipment. That's right. 24 hour shipping. Thank you. And then we've expanded that not just for ransomware, but for cyber recovery or disaster recovery.
24:40
And that even gets into a little bit of cold dr scenarios. Anyone here do Sunguard testing way back in the day. We're starting to go in a little bit of that scenario where we'll actually stand behind having these available to you and you not have to have everything ahead of time. So I want to give a little bit of origin story and then pull it forward,
24:54
please. No, I, I appreciate it. But, uh you know, there's a lot of these, a lot we can do uh around that, that guarantee, but very similar to all the others is about taking what may be a challenge if you have indeed talked to your legal team about what would happen in case of an attack and how that infrastructure is, you know, you can't touch it until the investigation is
25:16
complete and now you have nowhere to run your services from and he's trying to address that with just, you know, a commitment that will take care of it if that happens, right. I'll just call out a couple more things on this page just to uh to finish my round of, let me actually tell you what's behind the announcements. We have performance slas uh those are enhanced.
25:34
We have actually increased the guaranteed performance levels for uh some of our block and file tiers. So our Ultron premium tiers for those of you familiar with the portfolio have gone up in guarantees by 50%. No additional cost, no cost and more performance I know. I'm crazy. Crazy. Right.
25:52
Um You probably have seen that from pure before. I think we have a bit of a, a pattern of doing things like that. And um paypal and rackspace are people familiar in this room with the concept of the vendor paying you for the power and rack space, what it costs to run your infrastructure because we do do that with Evergreen one. So if you think about what Evergreen One is,
26:16
it's a storage service which you, in addition to being the user of it, very kindly have agreed to host for us at your own uh data center, right? Yeah, thank you very much for doing so. Um And because you're hosting for us, surely you need to pay for the power and the rack space that it costs to do that.
26:33
And if this is a true service, the vendor should be reimbursing you for that. And that's what we do as part of Evergreen one, right? So in Evergreen one, if you sign up for a um for a contract that again, we will ship you whatever infrastructure is needed to deliver to the guarantees that you have chosen. And with that will also come a check for the
26:54
power and rack space cost to operate that infrastructure, such that however much space it takes however much power it consumes that's not on you, that's on us. It can also sets the right incentive if you will for the vendor for us to make things more efficient over time. If we are footing the bill for the power, we have a pretty good incentive to make your,
27:16
your infrastructure better and better. Um Same goes for, for space. And also I just want to say it also shows how efficiently our arrays are that we could, it, it does help and this is I I think it's a common thread along some of these, these guarantees that we have in place. Why can't we do it? It's the evergreen architecture is the
27:37
technology we have, right? It, it would be a crazy idea to pay you for the power uh cost. When that power cost for competitive solutions may make up 50% of the bill. It would be impossible for us. That's not the case. How do we prove it by just cutting you a check? I don't want. It all goes a lot of this goes back to the
27:54
architect. Sure, like we can do pay power and rack because we don't consume so much power. There's a round of applause would cause out of 100 and 50 terabyte DFM. Today we do it like that. But at the same time how mu how many of people here will need 100 and 50 terabyte DFM? And can you talk to your cio about that?
28:09
Probably not, but you could talk to him about how the array is so power efficient that it takes less space and the vendor pure, your storage partner will pay for that because of that innovation. And the last piece I'd say is if you work with, um if you were to have a dual vendor strategy, I would encourage you to ask your other storage uh partner vendor if they could pay for the power and rack uh for their equipment as well.
28:32
We believe it is a relatively differentiated offer. I see, I see a question or comment there. Go ahead. I don't wanna pay for the high end performance guarantee on everything. Can you create an Evergreening contract that says, all right, you get X at this performance level Y at this performance level and Z we have here.
29:02
Yeah. And, and, and I'll repeat the question just for everybody's benefit. It's uh sometimes I remember um the question is around, you know, if we have maybe as a, as an internal uh provider of storage, you wanna offer a bronze, silver and a gold tier of storage. Uh How can we build that into an Evergreen one contract that gives you some amount at one
29:23
performance level, some amount or another, some amount or another? So there are 22 ways to think about that. Um One way is we absolutely have several tiers of servers as part of Evergreen one, we call these um performance tiers. The performance counter intuitively like one label performance is the lowest performance because above,
29:43
you know, nothing is slow at pure storage. Uh But above that, then we have premium and ultra, but these are really trying to achieve that. They're uh we can separate this into different licenses so that things don't fight with each other. We'll keep them uh absolutely separate and we'll make sure that all the highest
29:58
performance guarantees are on their own uh environment and we will scale that environment a lot faster as the needs grow. The second thing you can also do is if you wanna man manage that yourself, uh you can build an evergreen while one license for kind of the combined performance that you're gonna need basically do the the sum product of those things that you have listed
30:18
out and that will give you the additional benefit that in evergreen one, we don't limit each volume for you. We don't do what cloud does. So first of all, we don't charge you on provision space, we charge on consumed space. So over provisioning or thin provisioning if you will is is included. But number two, we're not gonna limit each
30:37
volume or each file system, whatever it is to perform to that sl A we're gonna look at it holistically on your total. We're gonna guarantee that the total level of performance for the amount of data you're consuming or have committed is sufficient. And if you want to manage that yourself, to give yourself a little bit of burst capability and, and you know, average it out across your um your various tiers of service.
30:59
That's a benefit that you get, keep, bring us home on this one. I'll bring us home on this one because I think there's just the, the last thing that I haven't talked about is evergreen 14 A I, I talked about it before, so I don't wanna spend too much longer here. But again, what is the, what is the uniqueness of evergreen?
31:16
14 A I? It's identifying that number one A I is very variable. So we wanna turn Evergreen one into something that gives you even more flexibility than a standard consumption offering, the more flexibility means that uh the the predictability of your cost as your data grows, but your performance doesn't grow because A I has a very constant performance
31:37
need based on those GP US, it gives you an extra layer of, of uh security that even though you don't have to size your capacity today, this will grow extremely cost effectively over time. And number two, it allows us to hit extremely high levels of performance in a cost effective manner for a customer. So what we do in Evergreen one for A I is
31:57
instead of saying, look, we know that most workloads grow with proportional capacity and performance needs. That's how a typical volume creation work flow goes. That doesn't really work for A I, right? For A I, somebody's deploying a whole lot of GP US that are very valuable. You don't want to keep those idle waiting for storage.
32:16
So you wanna make sure that you have a big pipe that you can run data through. So we're gonna guarantee that pipe, just agree on the price for that pipe. And then however much data you consume, you'll get basically cold storage prices for right. The idea here. And by the way, the minimum commitment on that cold storage is zero. The only thing you have to commit to is the
32:36
size of the pipe data. You can consume one terabyte if you want, you can consume zero. If you have actually not loaded any data on it yet. 11 minute before the question starts. So how many folks have GP us in your environment today? OK, good. Awesome. I will come back to h how many are looking to
32:54
in the future? We all know their supply constraint and they cost a huge amount of money. So if you have the most expensive thing in your environment that has been very hard to get and the most expensive people, A K A the data scientists having a service offering that is aligned to keeping the resource busy, the people busy,
33:13
hugely valuable. Do you have all the certifications and all that? Uh And the question is on the uh for government certifications to uh run this. So this environment just to be clear, still runs fully on premise, right? So this is not a cloud offering that we are
33:38
putting somewhere else where you have less control of your data. Um All of this still runs with the same control. It's an evergreen one offering. So you have the same control over your data and really the infrastructure that you have with any other pure storage uh platform purchase or subscription. Um So from a data security perspective, if that's, if that's where the question uh was
34:01
going, this is no less secure than any of our other products. It would map to the existing security certifications that we have from an on prem standpoint, this is the way to prioritize from a consumption perspective. So I'm happy to do a follow up with that, sir. You know, we can do go to the 120 that too and all.
34:25
But my question is that since you guys are involved in managing that environment somehow, uh do people have to be you citizen? Yes. Got it. That would be the question. Yeah. And, and I think this is really applicable to all of our green one. The question is, you know, the pure is doing more management work for this environment.
34:47
In addition to the product itself, we're actually, you know, keeping an eye on it and monitoring, making sure that you have the right gear in place at the right time are the all of those certifications in place. Uh Two aspects to that. Number one, yes, number two, the amount of management work we do we want to make it very clear that it's never meant to
35:06
interfere with the control you have over your data and the security that you have that nobody will touch your data without your consent knowledge and so on. Right. So for example, none of our uh monitoring uh that we do in order to guarantee performance and capacity levels ever looks at your data, we only look at how much of headroom do we have in terms of performance characteristics and
35:30
capacity characteristics. And then when do we have to give you a call and ask you for permission to come to your data center to install more? And of course, that work is done with cleared certified uh hints just like for any other install that we would do. Um but evergreen one is not about taking control away from you.
35:48
It's about taking responsibility of honestly just operational tasks that you shouldn't have to worry about is about offloading those from you. You can still provision and manage the storage with the capacity management, the upgrade. Absolutely. And, and very importantly, we don't get to touch the environment without your uh pre approval.
36:08
Let me, so we're gonna do questions at the end. We've interleaved a couple in. So it was like to be flexible that way. Um We have eight minutes left and I probably got a couple of minutes to wrap up. How much do you want to talk about? 100 and 50 Terabyte DFM S3 P three petabytes, 30 petabytes. I, I'm, I'm the, where do we get,
36:25
Larry? Come back in? You know, I'm, I'm formerly a hardware guy, so I'm happy to give him a shout out if that's OK. I, I think it really weaves in, into even evergreen one in a very clear way. And I think it goes back to what you said about talking to a cio about 100 and 50 terabyte DF MS or talking to them about the outcome and the benefit they're gonna get from it,
36:47
right? We're not gonna stop the development of really differentiated technology just because we're talking about outcomes, 150 terabyte DF MS, the denser flash ray chassis. All of these are things that we do to improve the density, the power efficiency, really the cost efficiency of our products.
37:05
And they're the key enablers for both our, you know, standard uh upfront purchase product portfolio as well as evergreen one. Um So the um density road map in particular where cause has shown that 150 terabyte module that yes, it is real. It exists, right? He teased the last year, I think now he's teasing the 300.
37:26
Um that is such a tease. I know um that is directly related to our ability to once again provide paypal and R pay uh payments as uh the infrastructure scales improve our energy efficiency guarantees and infrastructure scales offer more cost effective price points you've seen that with the E product family that we announced last year and in the ever group one
37:51
portfolio with a unified data repository tiers, more cost effective and lower performance uh balance tiers if you will for the use cases that don't require uh you know, the extreme performance that you see maybe on a flash or A X. So that technology continues to be one of the most important things that we do appear. Uh But we're trying to, as you've seen in the keynote today,
38:15
turn it more uh on its head and talk about why is this meaningful to you and this is meaningful to you as a customer only if it delivers a clear outcome. Uh And the clear outcomes are what we had in this presentation today. I think about how do you build something and get ready for it that's ready for A I that can actually go there legitimately, but actually provides value for you today.
38:39
So hopefully you're seeing that in fusion, it actually can have a lot of value immediately whether you're doing anything with A I or not, but it can set up the foundation for that. A lot of Evergreen one enhancements. This can drive you to as a service storage as a service platform internally and also set you up for when you're ready to go there. Because that's almost the biggest thing we're seeing right now is we're having frankly a lot
38:57
of A I conversations. What a surprise is a level of uncertainty and not knowing what's coming. How do you prepare for the unknown? You build flexible platforms that aren't rigid and can go different directions and don't have a high degree of, of, of lock in. It's on pure but lock in be from an on prem standpoint cloud standpoint,
39:15
manageability, et cetera. So that, that isn't a conscious intentional strategy for pure because we are in the very early innings of this. And if we picture like Gartner hype cycle in our head, you know, you can say where we are along that hype cycle, there's real stuff going on, but that's those are the waves that we're in and how can we help
39:31
you practically with that? I think before I go to questions, Larry or maybe not Larry, I wanna make sure you've been really good. What else do you want to add? I'll just, I'll just add another angle on A I infusion. So um to me there's really two the, the, the, the A I conversations in two
39:48
segments, like are you gonna run A I workloads in your environment? Meaning like do you need storage to do uh model train and inference and rag and those sort of things? And then the other aspect of A I is how can A I help us all in our lives and in our jobs, right? And like that's um that's what fusion is doing. So fusion is helping you to manage storage, using the A I workload planner and pure one to
40:12
do that. And then the other new thing we announced is the uh the A I uh copilot thing that is a um that's a very helpful addition to our portfolio you saw in the demo. If you were in the keynote, just being able to ask the system questions and get answers without having to know what Gooey was I supposed to go into? Where's the list of all these things?
40:35
How do I, I, I've got thousands of NFS file systems out there with export permissions on them. Are they all set up properly? How do I know where they all even are um to, to be able to just go to an A I copilot and say, are all my export permissions secure enough and that thing can go do all the work and figure it out. So I think that applies to more of us in the
41:01
near term as um we all try to figure out how we're gonna afford GP us and NVIDIA figures out how they're gonna make more GP us. Um So someday, so there's two aspects to it. There's running A I workloads and then how does A I help you do your jobs better and make your life uh Life simpler. Um We're all probably using Chat GTP to, of course, no one cheats and I,
41:25
and asks it to write their first drafts, but everyone helps it to use it to do the second draft. One last little fun thing there and then we'll, we'll wrap up and we'll stay around for some questions, but I wanna keep everyone here is even ch chat, chatting with, I think Prakash. Um So within the DX team, you heard him on stage?
41:40
I mean, we're, we're being careful with the private preview and even making sure that we don't have the challenge that some of their A I have A I Copilots have, which is being confidently wrong, right? So we want a level of integrity around what we're recommending to you because it's different than I'm asking it for the best recipe about whatever I'm making recommendations about what to do with your
41:58
storage and your data infrastructure. So there needs to be a high degree of confidence there. That's some of the timing. But speaking of uh previews, one quick last thing is uh um so a very, a very interim version of fusion and that remote storage management is available uh now in pure test drive, uh it's invite only. So you do have to request access to it.
42:20
So you can just reach out through your pure account teams and they will come and get a hold of me and I can create a voucher for you. If you don't know what pure test drive is, it's just a website you go log into it, you don't have to install or deploy anything, it'll provision a virtual lab environment for you with multiple arrays in it and, and then there's a walk through guide of step by step.
42:40
How do I create a fleet? How do I do all these remote um provisioning things? Um Our goal is to get all of our features available in uh pure test drive ahead of when they actually G A. So you can start to learn them uh and experience them. And so that's um available for anyone to ask about right now.
43:00
You cover the bottom two, talk to your core storage team about a test drive. There is a launch webinar on July 10th. Uh So there's a QR code for that on the next slide as well as so you can scan that if you like to attend the launch webinar, you wanna keep peeling the layers down of the, I like to think of the happy Young. You wanna keep getting more details on this as we go and there's actually a bunch of other
43:18
sessions here. Larry, ironically, yours was just earlier, but I think it was recorded. I'm pretty sure then there's also some other ones that you can attend. If nothing else join um myself and Rob Leman and J Wallace on Friday morning. Stay around on Friday. Please don't, you know, leave early. It'll actually be doing a,
43:33
a live version of pure report on plugs. So pure Report podcast as well as there's good technical depths with evergreen one infrastructure on the right way and uh deeper dive actually. Is that very top one? Was that yours, Larry? That's up there, but it already happened. So don't, don't try and go to that one unless you have a time machine.
43:49
And if you do, let's talk, let's talk about that. I think with that we are at time. Thank you so much for the time. Thank you so much for being here at Pure Accelerate on behalf of Larry Tago myself, all the pur, pure puritans and pure folks in the audience. Hope you have a great day.
44:05
Thank you. Thank you.
