00:02
I think hey I'm live. look at that. Cool. Welcome to what it says there. What's new in Flash blade if you were thinking and looking at the session builder, you're thinking like, man, that feels like a vague title or description. You're right because we're gonna talk about the stuff that was in the keynote that we couldn't
00:18
talk about ahead of time, obviously you're put in the abstract, you know, kind of thing. So thank you so much for being here. This is gonna be a little bit of a different session, um, actually I'm curious, were any was anyone here in the session yesterday. With Flash ray with Matt and, and Mayo, a couple.
00:32
Oh yes, Matt was in the session with Matt. That's amazing. So you know, kind of, yeah, so we're actually gonna kind of try and follow that format a little bit where it's gonna be a little bit more conversational. Uh, Justin and So I've kind of prepared a little bit to respect the audience, but we're not gonna be reading a script kind of thing.
00:46
So going through bouncing back and forth a little bit, we should have time at the end for questions, I believe. And if folks actually wanna shout something out in the middle, we'll kind of play a little bit of moderator facility. Here, so to that point, diving in for a little bit to,
00:59
uh, introductions if this is gonna go there we go. So my name is Andrew Miller, lead principal technologist at Pure, if you attended the coffee break series for the last couple years, you've actually potentially heard me kind of moderate and facilitate. This is gonna be that same style.
01:11
I just it on a couple times as well, you know, for the day job I cover the whole pure technology stack, but here today I'm probably gonna do a little bit more of adding color commentary and kind of hosting and moderating and for the amazing. Justin Emerson, Sona, do you mind introducing yourself? Sure. uh, Justin Emerson, I'm a principal field solutions architect.
01:28
I cover file and object solutions here at Pure. I've had a bunch of different roles. I've been here for a little over 5 years. Yeah, my name's Sanna. Um, you guys have probably saw me at the hardware showcase, same with Justin as well, but I'm a product manager at, um, Pure Storage. Been there for about 2.5 years working on
01:44
flashlight S&E. Um, yeah, looking forward to talking about the next generation scale. Awesome. So a little bit of a preview, not an agenda slide. It's a simple two part agenda. We're gonna jump with the time machine a little bit.
01:56
um, Justin and I were actually accelerating in 2022, and we'll kind of riff there and so I will be chiming in. And so talk about some of the origin of flashblades and even the promises that we made, uh, and even like not, not, not, um, literally the exact stuff that we said back then and then we'll dive into what is new this time you saw just a little bit I think it was in Rob Lee's session,
02:15
right? But man, that flash blade SR2 slide went by fast. So of course we're gonna dig a little bit deeper into that so. Just to jump in the uh time machine with me if you don't mind. So we were an accelerate in LA on the top of a parking deck,
02:29
I think, right, you know, take it away big parking lot next to, uh, next to a hotel in in downtown LA, uh, we was the first, uh, accelerate after, uh, COVID. Um, or maybe in the midst of certain parts of COVID depending on how you want to look at it and if we're still in the middle of it, but, um, you know,
02:47
that was, uh, the biggest, uh, accelerate for flash blades since it was first introduced because it was the first time we did a major, uh, platform. Transition. We went from our first generation platform, uh, which we launched in 2016 early 2017, uh, and, and came out with flash plates. And so, uh, all those pictures there, I managed to find someone in marketing who had all of the,
03:10
all of the, uh, the content from, from the Tech Fest 22, uh, and brought it up there. It was, uh, a nice little walk down, uh, down memory lane, uh, some folks who just walked into the room. You can see up on the slide up there. Well, actually I think you were, you were telling me the gentleman right up there is Tago,
03:25
Santiago and Vone. And there was this dramatic curtain pull like with the, uh, I think the flashlight exit earlier day and it was like in the keynote, but what you don't see is the blood dripping down his head. Apparently he actually injured himself so the blood, sweat and tears for Flashlight S was
03:41
both metaphorical and literal. And as well this was also a little bit of the launch we'll see there of some of the better science series. I almost wanna do a call out of that. Put better science, pure storage into Google or just go to the pure website and put it in. It's actually where we started to be really public about the underlying engineering
03:58
differentiation of what we do with Flash. You hear Justin with some podcasts around that other folks, I think I see Harry Harry in the Harry, Harry, I messed up your name. Sorry, you were, you were on one of those in the one of those podcasts with you. So we started to be very public about the engineering level.
04:12
Differentiation around direct flash and some other areas too, so right, right, and, and the big deal about Flash Blade S especially was that it brought that direct flash technology to Flash Blade, um, from, uh, where its origins in, in Flash array. So interesting, you know, story, right? So back when we launched Flash Blade,
04:33
uh, in 2016, you can see there on the left. Uh, there's sort of two boards inside the blade, and the blade was sort of the unit of scale and so it had both compute and storage in it, and that, uh, board there in the green is actually the original storage unit. You can almost think of that as the, the, the, uh, the prototypical uh direct flash module,
04:54
sort of the the proto DFM it's the Neanderthal of the DFM's Homo sapien. Um, and it is, uh, it was really a, uh, necessity to build the kind of things that we wanted to build, but with the, you know, advent of DFMs and taking that technology and sort of making it modular, it allowed us to then when we introduced Flash Blade S, um, separate the storage from the compute, um, which was very intentional because
05:21
what we wanted to do was set up the ability to. Somewhat connected to what we're gonna talk about today, upgrade the components independently in the future. So rather than having to replace an entire blade, which was both compute and capacity, be able to replace one or the other in a modular fashion.
05:38
There was also at the time too we were looking at. Even if you looked at SSDs at the time and how large they were and for what we wanted to drive for the original flash blade blades, this is the only way that we could get to the size. Yeah, it was, it was a sort of a, a necessity being the mother of invention in that, uh, you know, when we wanted to build at the time was like 52 terabytes of of man uh
06:01
flashing a blade. Um, I think the largest SSDs that you could buy at the time were maybe 1 terabyte, uh, maybe 2 if you were lucky, and so, uh, it wasn't going to be possible to fit, you know, 40 something SSDs in the space of that. So we essentially had to build our own and that was one of the sort of key unlock and insights. Um, that led us down the path of,
06:24
well, if we were able to do it for this product, we should be able to do it for all our products. That's what it's worth doing for all it's worth doing on the benefits. The one other pieces I like to toss them from a little bit of a history standpoint is if you look at the original flash play, this is back in vintage 2016, I wanna say it's kind of the design center when we were starting.
06:40
A lot of the applications that we were looking at were very high performance, very high throughput, high concurrency, etc. all the things that are great about flash blade, but actually a little little talking with someone yesterday they were talking about how they were an original flash blade customer, but all that data was essentially ephemeral. They threw out the data every time they ran it, so there wasn't any sense of flash play being a
06:58
long term archive or having data that lasted, kind of thing. So when we designed it originally we designed it for that application set that that's how you actually build that that's how from a project management standpoint you're targeting certain application sets and use cases as we saw the evolution of Flash Blade, it made sense for us to move to Flash Blades and so that we could get to the point of having
07:18
a nondisruptive migration, right? You already did that, but even without moving the data, so let's actually keep going. Anything else here before I keep going? cool. So let's think about the previous migration kind of not, so you getting from original flash blade to flash blades.
07:36
Walk through that, right? So, so Flash blades being sort of a in some ways a breaking change, right? Um, we had a new size chassis, we had a new blade form factor we had the direct flash modules, but right, we had made, you know. An evergreen commitment to our customers, uh, for the original flash blade,
07:55
um, I know Andrew, you've, you've always talked about things like the 1515 design architectural decisions, and one of those is non-disruptive everything we should never put our customers through a disruptive migration. And so, you know, it took a lot of engineering effort and huge credit to all of the engineering teams inside inside Pure, but, um, we were able to deliver uh a process we call platform upgrade.
08:16
Which was the ability to go from the first generation flash blade to Flash Blade S nondisruptively, um, and, uh, I've, uh, personally been involved in, uh, more than a dozen different, uh, specific customer interactions where, uh, we've been talking about how that process works and what the planning is and the feedback that I've gotten from, uh, customers has been overwhelmingly positive.
08:39
Um, I had one customer in Texas who they had a. Very, uh, complex high performance multi chassis flash blade system, uh, and they were able to nondisruptively upgrade that to flashblade S, uh, and they immediately saw huge performance benefits. They didn't have to take any downtime, uh, and the feedback was,
08:57
uh, almost I'll try and say this as almost verbatim as I can. That was the smoothest, uh, platform change we've ever been through, and it was so good that we want the rest of our stuff on Pure, which was great to hear. So there was actually a little bit of a look behind the curtain now for a second.
09:11
I don't know if anyone was looking in your schedule builder in the pure events app. You might have noticed that there were 4 people listed here, um, so we actually had a gentleman we decided to be, uh, we actually had, he had a he had a last minute family emergency so couldn't be here. But actually Christopher Munoz, this is where I'm gonna refer to my notes because I actually
09:29
wanna channel a little bit of what he was gonna say because he was actually gonna come up and talk about how he actually went through this exact, exact upgrade and it was fully nondisruptive so Christopher Munoz, enterprise storage engineer for City of Phoenix, only chose to pull off the slide just because he's not actually here and we're sharing on his behalf because we, we talked ahead of time.
09:47
He actually talked about how he had 10 years dealing with Pure. A huge environment to manage about about 2 petabytes and his flash blade, and I'm abbreviating I'm, I'm leaving out some of the fun facts about Phoenix City of Phoenix, by the way, which is that they have the it was the original canals back more more canals on Venice, yeah, yeah, by the way, so if you, uh, we were actually talking about if there's a,
10:07
if you're looking for a new business opportunity, you'd start doing gondolas in Phoenix because there's lots of canals you might need to be enclosed with air conditioning. I don't know, or maybe it's only a winter thing. Either way, OK, so back to the upgrade story, um, his comment was actually that flash blade to flash plates, they had a huge amount of data obviously, so you had to move that over,
10:25
but it was a 48 hour progress bar with no disruption and was way simpler than easier migration than previous migrations on other platforms. So even though we had to go in a scenario of flash blade to flash blade S for almost that kind of last migration ever, even that process was not disruptive for him. So channel. A little bit of that that customer story on
10:44
Christopher's behalf, yeah, and it, and you know then we called it platform upgrade because it it really was a nondisruptive platform upgrade. Um, the data was moving but it wasn't a migration in the traditional sense of somebody having to like plan things and schedule things and do cutovers from the outside perspective, the transition from that first generation to flashbla S was,
11:04
you know, from, from an application standpoint completely transparent. OK, so. Flashbla S was not a, it was almost if anything as much a foundation as anything. And so it was a foundation that then we did a couple things with it over time. So maybe a little more about a foundation. I think you're gonna be jumping near Tucson and
11:23
so go through the foundation in the last couple of years. Yeah, I can go through that. Um, so yeah, 2023 we were able to launch Flash plate E. So for here it was a cost optimized solution for a long term retention, um, archival use case which competes with HCD but better because of Flash. In 2024 we launched S-100 which was a hybrid
11:43
compete and it was an entry level model to flash plate S, um, which we were which has better I guess performance scalability than does um CMT we also recently launched too, which is a tiering data product from hot and um to cold um states which means that it would offer a lower TCO at the end of the day. And then most recently um we launched X as well which is a very high scale um AI
12:12
um I guess platform that we currently have. Yeah, and this is sort of reflective of the how how clever the design of Flash Blade S was is that because we built this modular architecture that had all of these different piece components, what we found very quickly was. Well, now we have the opportunity to use these same building blocks arranged in different ways
12:39
aligned to slightly different priorities and different markets in order to take that same architecture and spread it across all these different um areas. And so shortly after the launch of Flash Blade E in March of 2023, that was when you, that was that was your starting right when right when she when she joined, um, that is what gave us the ability to. Actually be able to address all of a potential
13:02
customer's data estate. So that was when Charlie really started talking about, you know, he said we have now a very different relationship with our customers because it used to be that a customer would always have to have somebody else, right? We would only ever be able to have, uh, be able to service a certain portion of our customers' data needs and now we really can,
13:22
can run the gamut. And the other thing I love about it is that we're ranging from all over, right? from. Like entry level workloads all the way to really high scale performance that we need to, so we're making sure that we have the right product fit for a lot of different new customer use cases. And actually, so I, I was in the keynote. I ducked out when Lynn Lucas was finishing up
13:39
so I could get over here, but it was actually kind of fun to hear Kas talking about, you know, that he's basically on a mission to, um, eradicate, maybe that was the word destroy someone who wants a job, but it was a it was a relatively passionate word about every single disc being gone. Except so it'll be a world of flash plus tape tape, but no one here really loves tape,
13:59
so that that's kind of fun although there's even a fun thing there we do have, you may have seen this publicly, we do have, we have made an investment in the ceramics company. Because if you're looking out maybe 5 or 10 years for very cold deep storage, um, it may be tape uh because actually tape is probably defied expectations over the last 10-20 years as far as density, but it may be ceramic as well so there's a little bit of a
14:19
fun exploration point if it's late at night and you're on the plane, you wanna Google on ceramic storage, OK. So Finishing up the Wayback machine with Flash Blades, we made a bunch of promises, and I think Justin you, you literally pulled this slide, so this is a this is a screenshot of one of the slides from 2022,
14:41
uh, so, so this, this is exactly what we said to customers three years ago, you know, when we launched the original flash blade, it was only offered with Evergreen, uh, foundation, although it did have some benefits of capacity consolid. that that um were were specific to that model but we weren't able to offer the same kind of uh colloquially colloquially known as free every 3,
15:02
the idea of being able to refresh your flash array controllers every 3 years we weren't able to offer that same kind of capability with Flash Blade because the data and the compute were were so captive, uh, to each other and so Flash Blade S offered us the opportunity to bring all of the cap. Abilities of Evergreen forever, formerly Evergreen gold to the flash blade portfolio for the first time.
15:26
So what we said was, hey, all the benefits that you've seen with Flash array, which had at that point been going on for, you know, more than 7 years, right? Um, you've been able to upgrade your controllers, keep your data in place, and now for a scale out system to also be able to say in the future you're going to be able to. Upgrade those nodes in the same way that you
15:49
upgrade your flash array controllers and with all of those other benefits of investment protection um that was the promise that we were making to customers that was one of the driving forces behind the uh architecture of separating that compute and that storage was to be able to do that in the future so it would be a pretty dramatic twist if we had to say we were, we weren't able to keep those promises, but.
16:13
News flash. The reason we're putting up there is that the point back that we haven't changed the slide. We're gonna actually walk through how we've been able to pull that off with SR2. Exactly. OK, so I think, I think we are done with the Wayback machine. Any, any final thoughts or at this point we're good on time.
16:27
So or if there's any questions that folks have from historically, feel free to shout it out, but we're about to go into what's new and cool and the building blocks and new performance question. So the question was, uh, zero move tiering is it available for S200? So zero move tiering as it is available today is a,
16:48
a separate solution that is a combination of flash plate S and Flash plate E in a single name space. And what it allows you to do is it allows you to. Uh, build classes of service and assign those classes of service or now, uh, be it habit driven be a policy to allow you to have two different tiers of performance but without moving any data around and so that's why it's
17:11
called zero move tiering is because it tiering is a solution. To a business problem, it's how do I, uh, cost optimize my, uh, my data and this allows you to do that by having this sort of uh hybrid system between our high performance flash plate S platform and our high density flash plate E platform but have those both delivered in a single name space and so it's not,
17:32
it's not. you can have. I don't have the number right now. Yeah, yeah, let's chat later. Yeah. Best performance we can offer. So we're taking the best foot forward.
17:59
having our yeah. Yeah, well, I think we've taken that feedback as well, so that's something we're exploring today as the S200. Well, you have the right people in the room here to have that conversation with. So maybe what what we'll do is we'll catch up afterwards and yeah it was literally gonna be
18:19
the let's chat about that afterwards because we want to go deeper on that if that's all right. So thank you. One more question. Last one you get. physical So, so the flashlight X, the metadata nodes are based on the same, uh, hardware as the, the S 500. Um, they have a specific configuration because
18:45
they're designed just for this one sort of metadata workload, uh, and I think we'll touch on this a little bit later, but as of right now. Um, ESA is sort of its own go to market, um, motion and so I don't believe we're doing any changes to the hardware in EA today, um, but if you're interested in, in learning more about it,
19:03
right, I don't think we're particularly bound by the performance of the S500 blades and an EXA today. Um, with the, yeah, so yeah, but what how that's gonna look right, um, the, the evolution of XA is gonna be sort of on a, on a parallel track to what we're doing with the, the, um, more enterprise focused, uh, flash blade leveraging a lot of the same building blocks
19:27
but going to a different level of scale than the classic flash blade customer bases looked for very much for AI, not that it has to be for AI, but long as that's where you see that level of crazy performance and scale. Yeah, good question. OK, so sorry, we are here to the last line. OK, that's I mean you can see that. OK, so let me,
19:47
a little more. Thank you. So thinking about where we are now. So let's think maybe a little bit about what are the building blocks if you don't mind that we use. So we're, we're gonna start, we actually debated do we start with the performance numbers and then work our way to how we pulled
20:02
it off? Do we start with the building blocks and then move to the performance numbers we're we're assuming that usually accelerate in these kind of sessions are a little bit more of a technical. Crowd, so we're like, mm, well we're gonna start with the technical building blocks and then work our way through the application impact.
20:14
So Sanna, do you mind kind of walking through the, the building blocks and any other commentary you have to to start us out? Yeah, for sure. I was just curious, um, are there any flash ray customers, um, in this room today? OK, so we have like about half the customers. So if you are flash customers,
20:31
I think you should probably know that this is the first time we're doing an evergreen refresh on flash blades. So we've been doing this for 10 years on Flash arrays, so it's not a new innovation life cycle that we're introducing on to this. We're just actually expanding it and making sure that there is consolidation amongst different products and introducing this for the first time over here and you might be curious
20:52
as to why does S200 and S500. Why not the other product models? Well, we launched originally in 2022 with that 200 and 500, and it was for high performance unstructured data which needed like high bandwidth as well. And so we want to make sure that we output the best of the foot out of both the evergreen refreshes and follow up with the other product models as well.
21:13
Awesome. That was a great save. I was supposed to ask you, how do you think about this at the highest level, and you did that, so that's good. Yeah, cool. OK, into the building blocks. I think that'll be great. Awesome. So when we're talking about the flash plate SR2,
21:26
what does that mean? Um, but before we even go into that, I think there's a few different areas we wanna talk about since we launched in 2022 of Flash blades. So for the flash with SR2, it's not just, uh, like, you know, blade upgrades that we're looking at right now which talks about the new CPU, new memory, which allows for like more power allows for more bandwidth,
21:46
um, we're also. Talking about the other pieces and um that you've also seen experience if you're a flashbic customer which is you know like new Nan technology for example I was talking to one of my coworkers and he was telling me about this one specific configuration about how looking at two of the same product models and different nan technologies it's about 32% improvement of performance of throughput.
22:09
And then for Purity we also have given customers back of like you know the the the benefits of like you know software feature upgrades also so some of the new things that we're unlocking because of our architecture as well for example we had Purity update where we provide um users of about 5% increase of like usable capacity and we gave that back to them because that's like something we unlocked later on so we wanna pass.
22:33
That benefit back on to the customers as well and I think Justin you had a few other software features you want to, yeah, and, and so you know there's there's all these software improvements that have been put into purity since uh the launch of Flash plays in 2022. A bunch of those are performance related, but a bunch of them are, are, are not necessarily performance oriented. They're more functionality,
22:54
you know, we, we mentioned things like CMT. We also have things like multi. Tendency and all sorts of other really important uh uh features that were um bringing flash blade into more and more areas of of uh sort of usable uh addressable markets um and so the combination of all these different things, right, some of this is the actual R2 blade,
23:16
uh, but uh a bunch of it is, you know, software improvements and that those NAN improvements. The, the flash improvements that, um, that Sona was talking about where you got, you know, a, a performance delta with our two blades and uh some of the newer DFMs that we've been shipping this year. Yeah, so just to itrate like it's just a continuous updates that we're giving back or benefits that we're giving back to customers
23:38
with the NAN as well as purity, but then for specifically for the R2 blades is the, you know, the 0.1 and 2 that we were making. In this case, as you would expect, we're incorporating industry advances because of course we, we should ingest those. The other point I wanna make I wanna go back to, I don't think I said this in the beginning,
23:54
uh, customer for 7 years, admin engineer, architect partner for 8 years. So I put my customer architect ad back on. I can only relook at platforms every so often there's way too much to do, so sometimes, frankly, I would relook back at something and when there was a hardware upgrade. There's a that's a little bit different with Pure because we're doing continuous software
24:12
upgrades uh throughout the entire life cycle more so than other companies that might try to major software advances to a new hardware. Platform. So if you're looking at this and thinking like, mm, it's been a couple of years since I looked at Flash Blade. I'll look at it again with SR2 pieces that you mentioned. There's been stuff around like SEC 1784.
24:31
There's pieces around multi-tenancy actually the previous one goes a little bit to worm. There's a, there's a whole bunch of additional efficiencies and features that if you have features 3 years ago that you're like, mm, it doesn't quite work for me. Make sure to take a look back not just for the hardware but also for the software advanced side. Exactly so.
24:49
Each time we so we like diagrams and we want to get a little more technical, so keep going. Awesome. So this is the actual blade that you get to see over here. So when we were when Justin was talking about desegregating the the compute from the storage, the compute being the blade and inside of there, and then in the storage being the two silver boxes you get to see at the DFMs.
25:09
So over here we want to talk about the DFMs themselves too, which is, you know, for GA we're offering 3775, but. And there's also we're going to be releasing this year 150 terabytes. So what that means is that you can scale from like you know, about like 260 terabytes all the way to 30 petabytes if you wanted to within a single
25:29
cluster. So that option of scalability, which also offers so much space efficiency over time as well, about how much percentage of space efficiency maybe like 70%, the mid uh mid 80s, low 800, around 70 to 80%. Exactly and not only that, but the overall architecture there's so many updates we've made
25:49
from the speed, um, standpoint like for example the new CPU so that offers more performance for like, you know, parallel kind of workloads which means that for example if you have like um two different type of workloads like we've seen in a customer environments right like genomics and backup, um, they've been really successful at these and that's exactly what Flash Ma is meant for for running all these different kind of uses use cases concurrently.
26:12
And um yeah so our two only helps enable and helps like you know further that narrative as well. The other updates we've also had was just like 100 gig, you know, gigabytes um of networking upgrade. What that means is that there's like you know even lower latency, high throughput. We have the TPM module which means that they're
26:30
secure boot enabled for customers when they're actually, um, restarting their system up that it double checks that the firmware is good for them to even run it. We also have more memory which means that there's more band. With um for more um I guess metadata operations and yeah I think that's mostly good from our architecture standpoint and how this is
26:52
basically the backbone for our performance and we can really like all the improvements and new innovation that we get to see it's all kind of starts from here and like you can't get the best in performance if you don't have the right architecture to support it. Yeah, and, and the other thing to comment on here is that this is the only part that's changing or that is required to change.
27:13
So yes, we're upgrading the networking to 100 gigabit, but going back to the original design of the chassis, the original chassis that we built launched with Flash Plate in 2022 accommodates that already. So there's no chassis change. There's no change in the fabric modules in the back of the chassis, cables, exactly, no new cables. We were talking earlier.
27:32
that this is gonna be the first evergreen upgrade that doesn't involve any cables because it's all gonna be front service it's pretty cool, um, but, uh, and then the just touching on the, the TPM piece you'll also probably hear about this, uh, in the sessions on XLR 5 talked about it yesterday talked about it probably yesterday in your in your other session, Andrew, but um this is part of our,
27:52
our secure boot initiative which is all about, uh, dev uh devsecops, uh. Able to uh make sure that, uh, we're, we're, you know, for, for especially for highly regulated or, or highly secure environments having this kind of, um, additional security, uh, signature verification of the code running on the system,
28:11
uh, is very important to a lot of customers. So we're bringing that now here with, with Flash blade with the R2 launch as well. It's actually a little bit of a fun little pure nugget there. We do share various resources from an engineering standpoint where we should. And we've actually been trying to do more of that over time with at a software level and a
28:28
hardware level so the secure boot and TPM capabilities are basically the same team deriving that yeah from a hardware engineering standpoint on both platforms which is a good thing for you and it's somewhat, it's somewhat, uh, um, fitting that Secure boot is coming to both XLR 5 and and SR2 at the same time and that both of those product refreshes are launching at the same time,
28:48
uh, because the original XL and the original. Flash blades were both actually designed around the same time in co-engineering and so if you take out a tape measure, you'll actually find that the dimensions of the flash blade XL are exactly the same as the or the flash array XL are exactly the same as the dimensions of the flash blade S chassis and they share certain power supplies and all sorts of other stuff.
29:08
So very cool, very cool that they continue to sort of be on a similar cadence. I do want to make sure to highlight and you hit this but just double down on that there's so much here that's not changing. And that's good because it's actually validation of the work we did in 2022 as we came and So and team came into the SR2 cycle, but they didn't actually realize,
29:28
oh man, these plans that we made three years ago, there's been enough change or we didn't see certain things. So that's good for frankly, continuity of vision, hopefully your confidence for engineering and ease of upgrade process. Both now and future generations, it's a validation of a lot of the design decisions that were made, you know, at this 0.5 years ago, taping it out.
29:47
It's OK, let's think about performance and maybe start from a generation over generation standpoint back to you so if I get it to advance, uh, magic there it is there we go. No, no, it's too much. I see the next one. I think I'm going to back to you. Yeah, awesome. So yeah, you might be wondering,
30:05
so when we first launched in 2022 to now, how much of an improvement did we get for performance and we saw 70% improvement in performance and that included the new Nan type, the new like software features that we've enabled, and also the new like blade that we actually get to experience as well. So all of those things combined together we see that much. Roman from the older version to now and that's
30:28
awesome because I feel like the when the customers that purchased them back in 2022 they not only are buying components in order for them to sustain longer but it's actually about not only just keeping up with the market but also being ahead of the market because we still wanna be competitive and make sure that you don't have to worry about am I still competitive enough we're doing that work for you and so right now we're climbing up to about.
30:52
Two times better than competition based off of the benchmarking work that we've done and it's an ever modern solution which means we're going to keep up with these, you know, the the evergreen tradition of replacing your hardware every 3 years for free if you're a forever customer. Yeah, there's, there's also some numbers we've been digging into around read write,
31:09
read IOPs versus write ops, you know this stuff matters over time and as we're working with you on individual opportunities, we can dig in at that level, especially in a public forum we're, we're seeing with the 7. but there's deeper data that we've been working on and can release over time or individually as we're working with you for application use cases and our,
31:25
our, um, our, our internal teams working on this have already brought this into into our, our sizing tools and everything like that. So we'll talk about this later, but the, the field is already, you know, our pure field is already equipped to talk to customers about what the, the performance numbers of, of SR2 will be for your particular workload.
31:42
Speaking of workloads. Uh-huh, the dramatic change. There you go. I do, yeah, so you might be wondering like what kind of benefits do we get in comparison for, like, you know, different kind of workloads like in comparison competitors, what,
31:56
what kind of benchmarking work did we do and how much are we claiming up to? So we have about 2 to 2.5% increase or yeah. For AI EDA and genomics comparison for S 500R2, so that's our highest performing con config that we have and basically what like what that means for AI HBC customers is that it's better or faster time for insights for EDA customers. It's like,
32:20
you know, faster design cycle, so it's faster, yeah, exactly, and time ties back. Money and then genomics is basically, you know, like the faster results you get at like faster diagnostic times that you can get as well. So time is essentially the one finite element that people can get and that's how we are helping to kind of optimize for them and you
32:39
can get more for less time. There's actually this is a total side note, especially on the genomic side, and we talk about gene about actually analyzing from a gene standpoint and we think about what that could mean for us at a personal level in the future around customized tailored medicine. When you see some of those claims in the news, this is some of the performances starts to
32:57
actually make that stuff more possible so kind of need to pull it all the way up to to a personal level, yeah. OK. So how do we get here? So, so, so how many, how many customers or how many folks in the room here are flash blade S customers? Hopefully see a few hands. All right, awesome.
33:15
Well, first of all, thank you. Uh, we, you know, always take a moment to appreciate and and thank our customers. Um, if you are an evergreen forever customer, this is already gonna be coming your way after, you know, you have your, your 3 years from, from your purchase of Flash Blades, and then if you do your 3 year renewal, uh, right, you know,
33:33
you renew out to 6 years at the start of year 4. You're gonna get your controller upgrades and so that means, um, the process for doing that is I like to say as easy as 123. So, um, if you have an empty blade slot, this is actually even easier because you just, uh, now this is obviously all going to be coordinated through peer support and we're gonna manage this through through our professional service but not casual,
33:55
right, right, um, so, so but the. Process is going to be essentially put in a new R2 blade, move drives into it, pull the R1 blade out, and then just repeat, right? So it's a very simple process. In fact, from a procedural standpoint, I was talking to somebody who works in, uh, field engineering, so an implementation engineer.
34:15
Um, the process is roughly the same as replacing a failed blade. It's from a, from a, um, from a process standpoint almost identical. It's just all blade out, new blade in and because all the data is stored on the DFMs and the DFMs are moving from the old blade to the new, you don't have to do any data rebuilding, you don't have to, um, you don't have a window. Where you're,
34:36
you're vulnerable to data loss, it's, it's really elegant and it and it's, uh, again, a testament to the design decisions that we made with, with Flash blate S. So super simple upgrade and, and like we said, it's gonna be the first evergreen upgrade where you can do all your replacements and you don't have to touch a single cable,
34:51
no cables, yeah. Yeah, and I think maybe there was a little bit too there about maybe networking and XFMs or. Yeah, because, because, no, oh sorry, go ahead. Oh yeah, I was just gonna say that that even for like, you know, the architecture we're talking about with the even the 100 gig gigabits that helps with all
35:09
the future programs that we're doing innovations and that's also helped enable us for um like I got September we're going to be launching something related to that but. In addition to that, we have like the new XFMs as well, which offers up to 10K file support system and it's we increase the memory from 16 to 32 so there's, it's not mandatory to actually upgrade that as part of this process,
35:31
but all of the other innovations and this goes back to how we're constantly improving our products, even if it's like not something essential, but we just wanna look out for customers and offer them more features. Yeah, that's a, that's a good point. When Flashlight S first released, we had our our XFM 3200s which were 100 gigs. Enable XFMs we now ship them,
35:49
uh, that are 400 gig enabled. So there there for example, if you want to be able to do more than 5 chassis or you want to be able to do more than 1.6 terabits because believe it or not, there, there's reasons you might need to do that now. Uh, you, you could, you, you can look at the XFM 8400 and now the XM 8400 R2, which are, uh, you know, uh, big, uh, bandwidth improvements over the XFM 3200,
36:14
but also not required to actually do. Are to upgrade, right? It's all cross compatible so you can have the original XFM, uh, along with, uh, the, the SR-2 blades with no problem. I almost want to channel a little bit of the keynote yesterday. Sorry, one second, uh, when I was asking, does anyone who out here needs 10,000 GPUs?
36:30
Let's chat. So anyone who needs more than 1.6 terabit, yeah, uh, hey, let's chat. I won't put you on the spot and make you we want to chat a little bit. So I we're gonna, we're almost through the end if you mind, I'll hold the question for a second if that's right, and then we'll do a question and thing, and you'll be first in line.
36:45
So, OK. So what I'm hearing is this is actually easier and previously was nondisruptive but even more nondisruptive yeah it's that's it's it's an even easier process than what we went through with platform upgrade and and the the customer experience with platform upgrade was almost universally positive and so this should be even, even more nondisruptive, even more. Simple to do and uh it's available today so
37:10
dramatic question. When do you think this is available? Anybody, anybody? OK, it is available today today that's available for both new systems as well as for like, you know, if you wanna do outside of the renewal cycle plate upgrades you can also do that from R1 to R2. So we are just about at the end.
37:29
So if we think about other sessions because we are, and the clickers are on a little bit, so actually, so there we have, there are other sessions where if you think about this, especially like you did one with direct flash yesterday, the sessions will be recorded and available later. So there's something you may want to go back and listen to. So I almost wanted to highlight there's some
37:45
previous ones that you may be interested in. You've heard teasers on today. Take us through some of the. Programs yeah, for sure. So I think that's the biggest advantage that we have with the Evergreen Forever program is that we will allow motions and give you credit for some of some of the motions that we have available. For example, capacity consolidation is one of
38:04
them where you can upgrade your drive size from, for example, 408 terabytes to 75 terabytes. You never have to worry about end of sale products with us. We will always make things. Either backwards compatible or like you know some alternatives which are even better for the customers and so that's one of the benefits that we are giving back but not only that,
38:22
if you for example purchase S100 or even S200 and you want to upgrade your blades, you can also do that too. You can upgrade from to S200 and S500 and we will give the credit back to you as well some some percentage of it. And and not only just like the free every 3 year program that we've been really promoting, we're also wanna make it sure that you also know that outside of renewal cycle if you want
38:45
to you can also upgrade your blade as well. The one thing we wanted to highlight here is you may be thinking, hm I see evergreen forever here pieces. What about evergreen one? Well this is the foundation, the architectural innovation that we do is the foundation
39:01
underneath the Evergreen one services, but there's not a direct correlation of, well, and there's not supposed to be hopefully if you're familiar with Evergreen One, it's storage is a service consumption based. It's not linked to a specific hardware platform, but the hardware platform. Enable that or in this case you might actually see hey, you could actually get the same amount of performance or more performance in an
39:20
evergreen one contract with a smaller physical footprint because that's what this enables kind of thing. So this enables and empowers evergreen one, but there's not a direct correlation in this session we're focused on the hardware refreshes and upgrades aspect of it. With that, I think maybe I didn't forget the question maybe uh maybe Justin and then Sana
39:39
final thoughts to bring us home. Yeah, I mean, I, uh, you know, I was, I was there at Accelerate 2022 and we launched this. It was a super exciting time for Flash Blade, and I am, uh, incredibly, uh, thankful to the engineering teams, to the product teams, uh, and, uh, thankful to our customers who have trusted.
39:57
Us and and you know put there in some cases most important data uh on our systems uh I hope you all appreciate, um, the, the part you know hope you know we appreciate the partnership and we're really excited to be able to deliver on the promise you know uh it's, uh, it's not, I'm not bragging about it because, you know, it's what customers should expect of a company like Pure,
40:22
but we did. What we said we were gonna do and uh I'm I'm really really glad uh at how well it's turned out. Yeah, I think Justin had all the points and I think the take home messages basically you know there's more performance, more security, more scalability and all like non-destructive as well and so um yeah I think that's the the kicking points for making sure that we,
40:43
you know, get those benefits back to our customer. It's all the things customers like about Pure. Some more future. I'm supposed to make sure to do see in the hardware showcase. It is there and then I think there's a make sure to sign up for your community.
40:58
Gardener Peer Insights. That's worth reading, by the way. You don't necessarily have to submit something there if you do like socks or you just like pictures of socks either way, um, make sure actually Gardener Pure Insights is a great resource for you to read other customers' perspective as well as please submit from a peer standpoint as well.
