00:00
Hello folks. Uh Good morning. Uh and hope you're having a great accel rate. 2023. Um It's a small room but there's a lot of seats up front. If you guys want to come up, you know, could have a more intimate discussion here. So, um my name is and I'm super excited to be talking about Flash blade and why it's really
00:21
the uh last uh scale out storage solution that you'll need for your unstructured data. Uh How many of you are using flash blade today or in the process of using flash blade? Awesome. How many of you considering flash play? I would like to learn more. Awesome. Great. Uh There's, there's something for all of you in this session.
00:42
Uh Today, we, what we will be talking about is why we built flash the way we built it. Um what we're doing uh today and where we're going next. Uh And rather than just me standing here and talking about how awesome our products are, we figured. Uh you'd love to hear from one of your peers. So we have Baron Judd Wave judges from Silicon Labs.
01:04
He's A R and D in uh infrastructure architect. Uh and they do some pretty amazing things. So we we learned from him as well. So, but before we do all of that, let's just take a step back and look at what's really happening in the market out there, right? We all know that unstructured data is exploding.
01:23
Um Enterprise strategy group estimates that by 2030 organizations will generate almost 10 x of the data that they have from today's levels just in a few years. And this is massive, right? Uh And so if you look at legacy on structured storage solutions, managing this kind of data growth uh is pretty unsustainable and
01:48
it organizations, I don't know in a show of hands how many of you face any of these, these constraints today? Anybody, all of you pretty much, right? You have budget constraints, you have, you know, complex siloed environments, security is a concern. Uh And that's not all all of this while you are trying to support your teams for innovation.
02:12
This is reality today, Ghana, the days of, you know, focusing exclusively just on growth, but that's not all right. Many large organizations are trying to chart a path towards net zero carbon footprint. Uh And as you heard, Charlie and others speak about this throughout this conference, 1 to 2% of energy consumed is by data centers, 1 to 2% of the energy that's generated
02:43
world over is consumed by data centers and transmission networks and out of that 20 to 25% comes from storage. And so it's, you know, so if you take a think about it, net zero carbon footprint needs to go through it efficiency. And this is a reality today, right? It affects bottom lines of many organizations. So what do, what does it really need from kind of unstructured data?
03:09
Right? Simple to manage, right? That's what really it's all about, it's about giving back your weekends. Uh And and future proof. Uh You know, this, like I said, this unstructured data is exploding. Many organizations are trying to keep data, you know, retaining data for a longer time.
03:26
Maybe they want to do you know uh a IML or training models on top of it. Uh So it has to scale with you without compromising performance, you should be able to consolidate multiple workloads right? Today, as you'll see in, in, in in a few slides, many of these are point products, right? So you need to be able to consolidate multiple
03:46
workloads security, you know, huge importance, you know, all of last year, pretty much every other week, there would be a ransomware attack on, you know, hospitals like state government entities. This is reality today, right? This is a constant battle, right? And this is not gonna stop and of course,
04:08
increasingly energy efficiency, this is what kind of like, you know, many organizations need from their unstructured data storage. Like I said, legacy infrastructures don't match up to this, they're not built for modern work clothes. Many of them are, have architectures which are maybe 2025 years old.
04:26
They're just continuing to just incrementally put stuff on it. Many of them are p products they uh strand data around uh and it becomes very complex to manage at scale and they're very inefficient in terms of energy usage as well as resources. Now, you could build unstructured data storage uh by putting a lot of commodity servers and scaling it out. But what you'll end up doing is creating a lot
04:53
more sprawl, right? Uh You could throw large scale disks like, you know, 22 terabyte, you know, or S SDS or anything like that, but you won't get necessarily the same performance or energy efficiency. And if performance is good for one kind of workload, it may not be great for another kind
05:11
of workload. So you can't really consolidate. So these architectures are not really built for that. What if instead you could go from all of this to just this, right? What if you can consolidate many of your workloads without compromising on performance or TCO or energy efficiency?
05:32
That's what flash lets you do. And so just like, you know, this is a big thing for us within pure right? We use um some of the principles of how modern systems are built and what you see on the left is, is really a very complicated, more modern system just like a smartphone kind of consolidates a lot of apps,
05:52
our design ethos is to consolidate a bunch of workloads and everything we do is really geared towards that. Uh That's what flash rate is meant for, you know, you can consolidate without compromise. How do we do this internally, we call this better engineering, right? Um Cos earlier spoke about uh you know,
06:13
kind of like the whole DFM and the purity software that's really kind of like the foundation of it, right? And he often describes it that, hey, the magic is really in the combination of hardware and software and that's how we think about it, right? There's a very famous quote by uh a computer scientist, a pioneering computer scientist, Alan Kay who says that if you're serious about
06:32
your software, you should build your own hardware. And uh this is, this is kind of how we think about it. Uh And so it all starts with our direct flash modules starts with PDF B. Uh This is what manages um uh you know, all the devices and, and does all the flash management and so on. But it's really just not about the flash
06:52
management itself. Everything we do is all tightly connected. Your evergreen story is based on this. Your TCO is based on this. Your longevity is based on this. Your energy efficiency is based on this at scale. If you're building 300 terabyte DF MS at one point, you know,
07:11
that, that's in our future. You can't do this, you know, just by taking off the shelf competence. You've got to engineering engineer this from the ground up and that's what we do here. And that's really our, our, our kind of like, you know, secret sauce if you will.
07:27
Uh but that's not all long life chassis. Uh We build chassis for an extended period of time. So, uh you know, there's the ship of this is like where you, it just, you just keep changing the competence and it stays current. That's exactly what it is uh blades, you know. So as you scale your environment,
07:46
you, you take your blades in and out and you replace them with new ones. Uh and your, your, your environment stays current uh integrated networking. Uh in many unstructured sort of solutions, you have to build your own, uh you know, bring your own network right to connect even the nodes uh in your environment. We don't do that, right. This allows us again to control the East West
08:07
traffic between all these nodes and that's super critical for us because that enables us to have predictable performance. And lastly, of course, cloud based monitoring, all of these together help you do the things that you can do with flash break today. Uh And, and that's what we strive to achieve, right?
08:24
This enables you to consolidate, it enables you to get the right tco longevity efficiency and all of that and so with all of these principles, we build flashlight s this is the latest, you know, this is the second generation, you know, we launched this last summer, you know, s stands for many things that you see, you know, but it's really a culmination of what we have learned over the years in terms of how
08:50
to build these large scale systems that can scale with customers. And it's a giant leap forward. It's really a game changer. It's got modular architecture, it's got unmatched performance efficiency. You know, in some cases, it's two X, some cases, it's even nine X in terms of our internal testing for things like metadata, iop and so on and it supports massive scale.
09:12
Today, we support a 10 petabyte system pretty soon we'll go to 30 pea bytes, let's like a three X within the course of the next, you know, maybe a few quarters and quickly beyond that, right? Uh As these new DFM modules come along, you can imagine that the scale just keeps growing exponentially and it doesn't consume more power or space or anything like that and it's
09:36
incredibly efficient. Uh And like I said, it's a game changer for us. But we also recognize, you know, we, we constantly talk to our customers. We also recognize that many organizations don't need all the performance that our flash data system offers today.
09:52
And many of the environments today are still on disk, right? 90% of the world's data. Uh you know, the data center still resides on disk, all the unstructured data piece. And we know that this, I mean, the whole theme of this conference is dis don't, right. But it's simply because it under delivers in
10:10
today's environment that's not sustainable. Takes up too much space, takes up too much energy, takes up too much resources at scale. Organizations are constantly like replacing disks all the time. I mean, that's really not sustainable, right? So we need to build long lived systems which can kind of grow with your needs.
10:31
And so we built a better solution. I talk spoke to you about the modular architecture of flash blade s it's built on the same platform and the modularity enabled us to quickly turn around and offer this. I mean, we felt it was the right time for offering something like this within a matter of six months. Um So Flash Lady offers uh you know,
10:51
the benefits of all flash, it's an incredible product, you know, it's delivered uh you know, lower total cost of ownership. Uh We mentioned this in our earnings call. It's got the fastest pipeline growth uh in the history of pure uh in the first quarter of the launch of any product. And that's simply because it's a very well built well engineered product.
11:11
Um It's got all the benefits of flash like I said, you know, power efficiencies, space and so on and so forth all for the price of. Uh and so if you're thinking about large scale systems for your environment, you should definitely look, look into this uh and both S and E uh you know, the part of the family now they're very power efficient and,
11:33
and we keep hammering home this whole power efficiency thing because we know that this is gonna become a big thing for large data centers. People are gonna start wanting to think about your overall TCO as your environment scales, you know, 48% less power, 28% less cooling required, all compared to E equal and NAS systems flash play D one watt per terabyte, 85% less E waste up to 1/5 of power.
11:57
Uh So, uh you know, these are built not just for your unstructured data needs, but also other considerations that your organizations may have for power and cooling and everything else and we serve multiple use cases. Uh Sorry, was that a question? Thank you, Miss Against the Yeah, we can, we can, we can provide that offline for you.
12:22
I I, I'll work with you on that separately. OK. Uh You know, we uh the, the question was is there a comparison between ens in terms of power and you know, they're kind of, you know, in, in many cases, um the systems are pretty much very similar, we use the same competence. So it depends on the metric like, you know, is it a per ru per terabyte.
12:39
So we got, yeah. So uh and, and so if there are any questions, please interrupt, it's a very small audience. So it's it's great to kind of like have these questions. So, so bring them on. Uh we serve multiple workloads, data protection, enterprise imaging packs, a IML analytics, uh you know your NAS systems, right?
12:59
So people always buy these systems uh for one use case and they constantly keep adding more as you as you learn pretty shortly. And flash blade s is great for the performance aspect of these segments. And Flash blade D will take care of the rest and it brings in the efficiency and, and TCO and capacity to many of these environments.
13:20
So that's really this family, right? It's the, you know, I hope you get a sense of why we call this the last scale out solution that you'll need for your own structured data. It's because we believe now with the introduction of Sne that we can cover a majority of the workloads in your environment, you can consolidate more and get the benefits of, of of all flash technology for your data
13:40
centers. Uh And, and so enough of me just talking about how awesome our products are. Uh I'd love to now call upon Judd uh Judd. Please please come up. Uh please ladies and gentlemen, please give a warm welcome to Jud Baron. Jed is R and D Infrastructure architect at Silicon Labs.
14:02
Um Jed. How's your accelerator been so far? That's been great. Thank you. OK. Uh You know, uh so tell us a little bit about silicon labs and what do you do there? Uh Well, silicon laboratories. Can you hear me? OK. I don't wanna be too loud.
14:16
Uh Silicon laboratories is a uh semiconductor design shop uh located in Austin, Texas all over the globe actually, but we're headquartered in Austin. Um It was founded in the mid nineties and we recently in 2021 sold off a part of the business and decided to become entirely focused on the internet of things. So we're, you know, making the chips that power, the remote control,
14:41
the light bulbs, uh you know, smart home or uh industrial automation kind of things. Um You know, one of our customers was in the keynote, you know, this is so that we had this morning. Um And so we're, we're doing interesting things with wireless, uh you know, low power systems owner chip.
15:01
That's great. So you build awesome chips, uh you know, power and kind of everyday home automation and, and so many other things that's, that's really cool. Uh So you must have a lot of unstructured data in your environment. What what does your infrastructure look like for, for, for kind of storing and managing this data today?
15:16
Sure. Uh Well, ed A shops typically uh the the data sets are very large and kind of sprawling small files. Uh you know, millions of files and directories uh for these projects. And uh you know, the, the whole design process, our H BC farms are, are backed by, you know, NAS NFS data storage.
15:39
Uh you know, it's probably, you know, 98% of our NAS storage is related to uh you know, this design workload. And um it's something that we've, you know, known for years is kind of the underpinning of the scale capabilities for our H BC infrastructure. And um you know, that that's, that's why NFS and you know,
16:04
fast storage is something that that's really interesting to us. Very cool. So you have billions of file, you know, millions and billions of files, like, you know, lots of directories. Uh What kind of challenges did you have before you, you became kind of like a flash blade believer. Well, uh so eed a workloads are, are very
16:22
heavily metadata oriented. So we tend to be in the 70 to 85% metadata from an IO distribution standpoint. And uh a lot of the storage vendors that we've worked with over the years tend to, you know, struggle like you have to do funny things with the configuration or how you provision your file systems, how you distribute them across the back end disks or you know how
16:45
you make, how much flash you mix in with it in order to, you know, get the right metadata ratios depending on, you know, what technology you're looking at and you know, with, with the flash blade, the thing that was really interesting when the product was announced for us was it, it's an all flash system.
17:00
And uh you know, it's, it's a aggregated meta data across all of the, the blades. So we were, we were immediately interested in when the product came out and then it was a question of uh you know, specific feature requirements that we needed. Um, you know, before we could implement it in a production.
17:17
Awesome. And you have flash data in production now. Oh, yes, yes, great. How did it change your outcomes for you? Well, in, in a few ways, uh so the, the, the slide decks that you, you put up, you talk about, uh you know, power efficiency or,
17:33
or space like we, we took uh I think it was 3.5 cabinets that were 52 U in, in height that were entirely full of like, you know, another storage, vendor storage, uh and basically got it into half of a 52 UK. So you went from 3.5 rocks to about half a rock. That's right. Pretty much. Ok. Yeah.
17:54
And it got faster, of course, which was the other thing. So it's using a substantially less power because, you know, we, we, we yanked out, you know, I think it was something in the order of um, 1300 dish spinning disks. Uh I, I think that's right. And, and you know, went to, you know, half of a rack of,
18:14
of flash and that's, that's in our, our large uh you know, ed a data center here in the US. We have another one that we set up uh overseas that as soon as we did it, we were just like, OK, we're gonna, this is all flash, we're just gonna buy it. Just started with all flash that OK, That is awesome. That is awesome.
18:30
Uh You also, we also spoke about consolidation offline, like tell us a little bit more about what you do for consolidating. Well, so other than the ed A workload, which is, you know, the, the the, you know big rock, so to speak from a storage perspective, we also have a couple of other workloads. There's, you know,
18:46
just general unstructured data type SMB windows workloads. Uh So there's some of that. Uh we have some uh you know, kind of B to B style file dropbox stuff uh for some of our business applications that also use uh SMB protocol. And then we, we have some workloads that were um you know, related to manufacturing with like parametric data where we collect uh test data
19:13
from the chips that we're manufacturing and you know, that ends up landing in storing and used by multiple systems that are looking at um kind of how our yields are and, and things like that and, and they're aggregating that data and looking at it at in scale across, you know, all all these different ships. So you know, having that work fast is, is good for our operations team.
19:35
Very cool. So you went from 3.5 racks to half. In addition to that, you were able to consolidate a lot more workloads in the same system that's while improving performance. So that's like, you know, like we spoke about consolidation without compromise. So that's great. So tell us, where do you plan to go next in
19:51
your journey? Uh You know, and with flash blade, well, at the moment, we, we're not uh we haven't, we haven't globally swept uh you know, to all flash blade, but it, it's, that's, that's the, the next kind of immediate thing that we want to do, uh you know, for standardization purposes.
20:09
The other thing that's really nice about the flash blade is, you know, we, we get great support. Uh We, we don't have a lot of issues with having different vendors have to come in to do, you know, disc replacements and stuff and having to, um you know, I don't want to jinx it. There's not any wood around here to knock on,
20:25
but uh you know, so I feel like, uh you know, we, we have a lower number of failures. Um And, and the, the support is is excellent. So um being that I don't have staff on my team that manages all this infrastructure at every site that we have it. Um you know, where we, we have NAS storage, that that's a, you know, real boon for us if we can put something out there that we know is gonna be
20:48
reliable and, and have um you know, excellent support. So that, that's another one of the driving reasons for us. And then, you know, we also been working with pure on a number of things I, like I mentioned earlier when we started looking at flash blade, there were certain features that we needed, right?
21:05
Like NFS four with access control list support. Um As we continue to uh you know, engage with the folks here at, at here from a leadership and product development perspective. Uh you know, we're, we're trying to, you know, basically execute the same thing that they talked about with like incremental gains, trying to find ways to continually tune the
21:29
speed of our H BC farms. So we can hopefully get to market faster with chips. There's a, there's a significant lead time from like, hey, we've got a chip that we want to go make. We're gonna put some engineers on it. Do. The chip design is complete and they're gonna send it off to the foundry to be manufactured to, we've got that te silicon back and software
21:49
development is happening. Firmer development, all that and then it's available for, for customers who got design wins. We're shipping and volume, anything we can do to, to shorten, that really gets folks like the CTO and, and you know, the head of sales, they get really excited about that.
22:02
Yeah. So, so performance and reliability and enabling you to kind of like, you know, move faster uh to get to your outcomes quicker. That's what you look for in your environment. Exactly. So that's awesome. If there's anything else that you'd tell the audience in terms of working with pr and what, what to look for and working with flashlight,
22:19
what, what would, what would that be? Yeah, I mean, I, I've been working with Peer since 2013. I think that's when we bought our first. Um, yeah, you know, flash array and, you know, that immediately made, uh, you know, waves with, with that platform when it came out. And, um, you know, I've worked with the account team in,
22:38
you know, Texas, uh you know, throughout that time up and, you know, when Flash Blade was announced and, and multiple times throughout that I've been engaged with, you know, leadership folks such as yourself. Um, and, and uh other product development, people at Peer over the years and they listen and, you know, a lot of the other storage vendors that I've worked with in my,
23:01
you know, 26 some odd years in information technology, don't listen to all their customers, they listen to a handful of really big spenders, right? Um And I appreciate that like we, I feel like we have a voice that, that gets heard. Not just because I'll come and sit at a conference and talk to people, uh, that they, they actually will listen to you and,
23:23
and put stuff on the product road map and prioritize things and, and that's massive because what I need and what my industry needs, you know, isn't always considered when you start talking about certain other, uh you know, companies that we've worked with over the years. And I mentioned the support thing again. It's another just,
23:43
um you know, amazing aspect of, of pure, like I, I feel like over the years when you, you pick up some new storage technology or servers or whatever it is and there's the, the new kid on the block and they're doing something disruptive, they may have excellent support when they start, but 10 years into it, you're like, you know, if anybody has worked with VM Ware,
24:06
I'll pick on somebody outside of the storage stuff, you know, their support has gone from great to abysmal in my opinion. And I, I would have to say that I don't think that that experience has happened with pure. At least not yet. I mean, hopefully it won't get that way.
24:21
Uh you know, things like that, that, that make a big difference for us. Yeah, look, yeah, I mean, from our perspective, you know, we think of, uh you know, our customer relationships or partnerships, right? We want them to be successful, you know, only then we can be successful.
24:36
Uh, and so, uh, you know, listening to big customers, small customers, everybody in between, uh, really helps us, uh, you know, make the product really available for everybody and, and that's what we're all about. So, um, you know, thank you for that and, uh, Judd have a seat, uh, please don't go anywhere. So, uh, we'll wrap this up and then,
24:56
uh, quickly and then we'll, we'll get to Q and A. Thank you so much. Thank you for your time. Appreciate it. All right. So you heard about how awesome flashlight is in, in a real life environment. Uh And you'll see many, many customers like today talking about flash.
25:11
So please seek them out and, and learn from them. Uh And so let's talk a little bit about what's next for flashlight, right? Um Clearly, you know, we are relatively new compared to many of the other vendors in the unstructured storage space. I'm saying relatively new because they have a long uh you know,
25:28
legacy architectures and been around for a long time. So there's a lot of things that we have to do ourselves, right? Uh For us, you know, there are three pillars uh from our perspective in terms of how we think about the future. Uh clearly our hardware platforms are there. Uh Now we have to continue to scale the core platform to make sure that we match customers
25:46
expectations. Both in terms of performance and scale and consolidation uh and everything and, and grow along with them. So that's a big focus for us. Uh sorry, broaden enterprise reach. Uh you know, this means that, hey, making sure that as uh customers, our customers themselves become very big.
26:02
Uh We also uh you know, talk a little bit about, you know, make sure that we are uh giving them the tools necessary to do the things that we, we spoke about, right, enhance security, you know, better consolidation, better knobs for controlling workloads, all of those things. And lastly, uh you know, in the unstructured data space,
26:21
it's a, it's a long tale of very uh every vertical would have very, very specific requirements. Your media and entertainment vertical may not be the same as genomics, they'll all have commonalities for sure, but there are kind of nuances across the board. And so we want to make sure that we address all of those little things uh you know,
26:39
in, in each of these use cases. So that's a big focus for us. So what are we doing today to address those? And, and what will we be doing in, in, in the future? We spoke about flash play D and how great it is. And we've launched that just now, just last month, we've already started getting a lot of interest in it and we will continue to do that.
26:56
Um SMB three is a big focus for us. We will launch that later this year in early part of next year. SMB is one of those things that hey, there are so many different quirks and how it interacts with NFS and especially for many unstructured data workloads NV GP direct storage is going to come out. Also end of this year. A lot of our customers have said,
27:18
hey, flash pay is awesome. It's great. But I'd love to learn more about what's inside, like, you know, give me more diagnostics, more metrics. So that's a big focus for us to kind of expose many of the internals that we capture in a consumable format. We've for the longest time shied away from
27:33
throwing a lot of data at customers. That's been a design focus for us even on flash array. And the reason for that was we never could figure out why customers would want a particular metric. You know, it may just be comfort factor, like say, for example, um 100% on CPU, right?
27:51
The storage is constantly doing stuff 100% on the CPU on one blade doesn't mean much because it's actually doing work and you want it to run at a, you know, a very high level and things like that, right? So, but, but we will constantly look for ways to expose meaningful metrics so that leak to customers to take actual action, right?
28:09
Uh and not through the kitchen sink at them. So, so that's a big focus for us. And lastly security, uh we want to get into more of the federal environment, really large scale uh systems. So security becomes a big thing, especially with all the regulations and so on. Uh So that's a big focus. Logging, auditing is one example,
28:25
but there's so many little things that we do uh to, to really focus there. So that's what we're working on right now. That's the next six months and maybe the next nine months. Uh And where are we going beyond that? Right? A big focus is of course, leveraging our expertise in flash, we have 10 years of experience taking uh you know,
28:45
flash that was built for the commercial market and converting it into enterprise storage, right? Uh And so as you can see the the gap between HDD and uh you know, the DFM technology that we will have will only continue to grow. And as a result, the TCO differences will, will, will only continue to be to favor us and our systems. Uh And so we will have bigger and larger denser
29:11
direct flash modules and so larger systems. So what does that mean? Really? Right. So we spoke about flash as a whole. So 30 petabyte systems will pretty much be around the corner. We'll start with flash blade D and then we'll get to flash blade flash D because this will enable you to have large scale capacity systems
29:30
for compliance or long term retention. All of these environments, right? Uh data mobility like now that we have two systems and ask question is, hey, what are you doing to kind of like move some of the data around? Right? So we're working with multiple different things, you know,
29:45
just, just thinking about how we can solve this problem. So that's a big focus for us. And then of course, you know, R talked about cloud operating model. It's really about how do you consume storage the way you do it on the cloud, which is orchestrated, automated and really, really simple and easy to consume, right? Either as a service you pay as you go or
30:05
various different models, right? So fusion is our product for doing that. And today fusion supports flash array and pretty soon it will support flash blade as well. Uh So you'll get the same uh you know, benefits of cloud, but on prem, right? So those are some of the things that we're
30:22
working on. There's, there's more things that we can of course talk about and share and if you guys want to have uh conversations offline, I'm more than happy to do that. Uh But that's all I had today in terms of prepared material. Uh But that we'll just 01 last thing. OK. So this is what we spoke about and I'll just
30:36
quickly summarize here, right? Consolidation of multiple workloads we spoke about that. That's what it customers need. You know, Judd also gave an example of how he's doing it in his environment, energy efficiency, why it's becoming more and more important,
30:49
scalable future proof architecture supremely important for the unstructured space, right? And of course, has to be simple to configure and manage. Judd gave an example of, hey, he doesn't have a large team to manage systems across the world. So making it simple and easy for somebody like him so that it it takes away the burden of just managing the storage.
31:08
So you can focus on other things very important and lastly security, right? And all of these things are stuff that we work on today. They delivered by flashlight and we'll continue to do this and with partnership with you, our partners and customers, we hope to continue to do so in the future as well.
