31:35 Video

Creating the Unstructured Data Advantage with the new FlashBlade//S

Watch the introduction of FlashBlade//S—the first and only scale-out storage platform that efficiently powers your modern unstructured data needs. See how the new FlashBlade//S family eliminates traditional scale-out storage limitations, allowing you to achieve simplicity and the multi-dimensional performance required for consolidation of key unstructured data workloads.
Click to View Transcript
00:00
It's really good to see you all. But it's also a little crazy to see like the past five years of our professional lives condensed into 45 seconds. So little family story since we're, since we're on that path here, I don't have any pictures, but I have my my five year old turns six on on Friday. And um, he was one when we launched Flash Blade. So that's how I that's how I kind of gap the
00:23
time. So kind of no sleep launching the first version. Getting a lot of sleep now, but you know, same, same, same stress over launching this new thing. But man, is it gonna be fun? It is gonna be really fun because when we think back to five years ago when we introduced flash blade, we were making some pretty big bets. And actually a lot of you told us repeatedly, we were making big bets often in the form of,
00:44
wow, that's really bleeding edge. And are you sure it's not a little ahead of its time because we were betting that unstructured data would not only continue to grow of course, but that it would demand increasing levels of performance. We were betting that fast object would become table stakes, that it would no longer be relegated to the slow, cheap and deep archive and that
01:03
applications ranging from Splunk to calm vault to snowflake would want to take advantage of the simplicity of the S three protocols. We were betting the ai would go mainstream, especially in the enterprise. And we were betting that simplicity would matter as much, if not even more in the scale out unstructured data market versus where we've seen it play out very significantly in the block market.
01:26
Most importantly, we were betting that our customers would benefit from being able to cost effectively address a diverse set of technical and business challenges being thrown their way, and that we could deliver that through something that we started to call. Like better science. Better science. Uh look, I think better science is a pretty tongue in cheek.
01:45
Uh, you know, kind of phrase, um, charlie hit this in our in the in the earnings call. And uh, it really is, you know, looking at the industry who are sexually building products on commodity components. We're not building things on commodity components. We're building high technology that drives value. And so, you know, when we talk about better
02:04
science, we kind of mean, you know, what does it mean to have a metadata structure that can essentially scale infinitely to or to hundreds of billions of objects, right? Or eliminate the overhead associated with an off the shelf SSD these are all things that are major benefits as a software architecture. You know, that takes advantage of hardware technology.
02:25
We're bringing the hardware up near the head room of our software. Right? That's always been pure mantra is we have software that sits here, we've got to bring the hardware up to here, commodity won't get us there, commodities won't do it. So when we think about, you know, what that means for,
02:40
you know, kind of the lifetime of a system. And I'm really pleased that charlie that charlie touched on this this morning, um I wasn't, I hadn't seen him rehearse, so I didn't know exactly what he was or how long he was gonna be talking about it. Um, but oh, wow, that was actually kind of funny because you did go along just to be clear, like good job charlie gives everybody else a hard time for going wrong.
03:01
So charlie charlie went along to. Um but you know, the point of this is is that we are on that thread of sustainability and thinking about things in terms of of of of decades and and usability and reuse and repurpose as opposed to, you know, throwing things away and and not getting use out of those things. So what do you think you think there's any of this better science innovative high technology
03:25
somewhere here in this tent that we could show everyone? I would say there's a lot of it, but there's one in particular that I think most people are here ready to see. So how do you wanna do this? Well, let's roll the video first. Okay, we spend so much time on the unnecessary, the unknown.
03:43
The unruly what we need is the uncomplicated flash blade? S it makes complex data storage simple, eliminating chaos, combining hardware and software innovation, decoupling capacity and performance, while delivering power space and cost efficiency
04:20
evolving over time to meet your needs now and in the future. This is the last scale out storage platform you'll ever need. Okay, so, so I would say, here comes the trapeze act of this, of this, this part of the performance, right? Like this is gonna be, we gotta see if we can pull this one off, how we, how you want to do it.
04:48
Well, so, I mean we want to tell them where it is. Yeah, but we don't all leave this room and go out and check it out. Yeah, so we're gonna tell you guys where it is. Do not go out and look at it. There will be an opportunity to go out and look at it. You will be able to see it, touch it, feel it, pull the DFM talk to to go and I'm Rita,
05:03
they will be out there cause they're awesome and they can, they can help you see it and understand it, but we're gonna find a way to celebrate this. Let's say let's do a countdown countdown, everybody with us. Yeah, Okay, let's go down. You Guys Ready? 3, 2 1.
05:20
There it is. Well, I can't dance. Look at that. Look at that. A lot of work went into that thing, a lot of work went into that thing. So look, we're super proud of this. Um and there's a lot of people that that contributed to this. Um, but what makes, you know,
05:53
kind of Flash blade? S so special is what I referred to earlier. It's hardware that meets the capacity of the software and vice versa. And it's a, it's a philosophy that shared by, you know, several, I think pretty smart companies. Yeah, I mean, I'd like to think that we are in good company with having this approach.
06:12
We see these other organizations that deliver truly exceptional products and uh, you know, we all kind of seem to share a common way of thinking and they were kind of missing the slide, but the companies, but those companies would be Apple and they would be a sure they would be the cloud providers. It would be Tesla, right? These are companies that, you know, are
06:29
marrying hardware and software to make good things. So there's our slides, so I don't need to use props. Oh, it's slides back there. But the slides not up here anyway. If you can all turn around for one second and look at that one monitor, we can make this work. Uh, we got it now.
06:43
Thanks guys. Okay. So we didn't just decide to start with this approach as we started building the flashlight test product though, did we? You? This is the result of a lot of good technology decisions. Um, that, you know, our founder cause made early on,
06:59
who had this belief that if you could rewrite the IO substack above nand instead of just repurposing the one that you wrote above disk? I know, I just lost some of the people in the room, but I definitely got some of the other people. So that's great. I'll take the technical people right now, everybody else you can kind of tune out for a second.
07:14
Um, but you know, those choices then bled over to people like rob lee and brian gold, who really were, you know, the sort of originators of this flash blade flash blade product and you know, one of the key decisions was not to use off the shelf SSD s. So, you know, that allows us to um, you know, and have internal controllers allow us to manage kind of each drive, right? So instead of having the overhead with a
07:39
packaged SSD, we've taken all of that out of the mix. And what that allows us to do is it allows us to use the same direct fabric modules or DFM because we are acronym people, I have to remind myself to say direct fabric module before I say DFM uh, DFM technology from the flash array. This is a big, big seminal moment for us. We're using the same Q L C drives from the
08:03
flash array C uh, that, that are now in, in, in in the flash blade s and so what that does is it unlocks about 20% more capacity from the nand and I want to think about that right? 20% can kind of uh, is that a good number? Like 20%. 20% of an exabyte is a big number and that is the direction that we're heading. 20% of an exabyte.
08:27
4% in exabyte scale is a monstrous number. 20% is huge. So think about this today, 30 terabyte Q L C enterprise SSD s or what ship in the market in the flash blade? S We are shipping 48 terabyte drives for per blade up to 10 blades. Right? Not only that. Right, This is kind of that point that charlie
08:51
was referring to earlier, you can take the path, you can take the commodity path and you can put hardware in there or you can take the software path. Which one do you want to do? Do you want to do the work to create the right thing or do you want to create the best thing, the better science thing or do you want to find the way that just gets it out fast
09:16
by the way? Guess what? Neither one is wrong. One is just better. I mean, if you really sum it up, it's effectively that our hardware innovation and our innovation, our hardware software co design. It's been really focused on two key things from a flashlight perspective. Right?
09:33
The first is exabyte scale exploits eventually but also unmatched efficiency in terms of power density and performance here charlie talk a lot about that this morning as well. When you think about it from the perspective of potentially needing to have your storage systems support up to hundreds of billions of objects. You understand why you would need a nearly infinitely scalable metadata architecture
09:55
underneath. And when you're talking about efficiency, what you don't want is to need a big layer of SCM two front end, your Q L C. Is kind of like a uh, cashing tier just to pin your metadata and get the level of performance and resiliency that you're looking for. This efficiency is you keep hearing us talk about is one of the things that we are most
10:14
proud and excited about in flash blade. S and in fact it's actually the most efficient scale out platform on the market in terms of the sustainability metrics that matter most to you. Terabytes per what gigabyte per second, per watt, terabyte per rack unit. And we're gonna only continue to enhance that advantage over time Now, just to get a little bit more detail that kind of touched on this.
10:37
But uh, you know, a single chassis can support up to 40 48 terabyte drives. That's how we get to up to nearly two petabytes and effectively comes in two flavors. The S 200 for ultimate efficiency and the S 500 for extreme performance and we're delivering more than twice the performance, density and power efficiency of the original flash blade. So it seems like we're pretty aligned to
11:00
charlie's messaging about innovation. It's like we've been listening to our Ceo Ceo, who would do that? Favorite aspect, Favorite aspect of flash played us. Definitely the modular desegregated architecture. Oh, modularity modularity matters. Uh rob lee is in the front here smiling laughing because we've had multiple debates
11:19
about modularity. Um, and look, if you think about what we've done, we've disaggregated the storage from the compute right? So that means that capacity and performance can be upgraded independently. So let's think about this in the reverse. If those two things are married, your costs are always going to go up because both things have a cost to it.
11:38
What we've done is we've made it so that you can you can customize your platform and your configuration to tailor it to your workload. If you need capacity, add capacity, you don't need to add performance if you want to add performance but don't want to add capacity, we can add performance. That is very, very, very key. And critical component to being able to design
11:59
what you need, not by more than you need. So with every component being modular, we move much closer to what I think some folks here know of the flash of a model. So our R D F M S R Cpus are networking. They all leverage the latest innovations in the industry. And what's important is that they can leverage those innovations as we move forward.
12:23
Non disruptive li in the chassis in a modular in a modular fashion. So, you know, the core principle here is a move towards modularity and that is, you know, what enables by the way, Evergreen. And I say this every time I'm on stage, everybody's heard it before, I'll say it until I know it's understood by every person on the planet. Evergreen is not a business program, it is an
12:46
engineering philosophy. It is that which enables us to get to the sustainability claims that charlie made this morning right? Like we have a social and moral obligation, I believe, to be more efficient in how we use things like power charlie hit on it, we're gonna we're gonna pass the world off to our Children,
13:07
that is not the world that was handed to us. And so I view this sort of passionate desire to get to modularity as one that does actually in some way tied to the environment. And I think that's a good thing. So I think you can really see this flex flexibility and modularity from the image up here. Um you know, you heard we can scale from up to
13:29
10 blades in a single chassis, up to four direct flash modules or DFM per blade, but you aren't tied to a specific performance capacity curve for your workload. We know your needs can change over time and we want to provide that investment protection so that you can start with 7 24 terabyte direct flash modules as the smallest configuration in that chassis and and up to again that nearly two petabytes and the two petabytes.
13:54
What almost two petabytes raw in what capacity? Single chassis and how big the single chastity. Five you everybody got that two petabytes of five, you know, server around it, servers are in the blades, looks like that looks like that, you can go see it later. Uh, so I think the most important thing though is that this modularity ensures that it can evolve with your organization's requirements
14:13
over time. You don't need to as matt talked about, you know, totally anticipate exactly where your requirements gonna come from. We can continue to grow and evolve the system to meet your specific organizational needs. Okay, broken record time Because I've been a broken record on this for going on my 12th, I'm approaching my 12th anniversary of your, I was, I was recently told that that's not how
14:34
other companies work, but apparently every three years big storage shows up at your facility and says, hi, can I have your frame back because I'd like to tell you something new. Right, Well I don't think that's actually the best way to, to, to push push this. Right. If you think about the way that the flash array model has run. Right. And, and I know that evergreen and flash array,
14:55
we have customers in this room that have enjoyed nearly a decade of non disruptive evergreen upgrades. Think about that. Think about that Bill Gates said and I'm paraphrasing, he said we all expect things like self driving cars in two years and were disappointed when they don't happen. When we look back at things from a decade.
15:18
Were shocked at how much changed. Think about the think about the person that runs storage in an environment that has had the same virtual serial I. D. Number or has had a system that's gone from a dual control or modular shelf expansion architecture to the flash array Excel today That is 10 years of evolution and maybe that system is the same size or smaller.
15:43
It's just faster and better. That's pretty that's pretty incredible. And the flash blade? S architecture delivers similar capabilities for us now on unstructured workloads. That's why I like to say uh you know other storage uh gets worse over time. Uh and the truth is our storage literally gets better over time. Who gave you that line?
16:05
You did? So you should explain it. I did, I did steal that from amy so Alright so well our architecture is highly differentiated in terms of delivering a scale out unstructured data platform. What really makes flash played as a game changer is how we leverage that architecture to deliver rapid and ongoing improvements to our customers.
16:24
And so with Flash Blade us we can support all of the all of the evergreen elements so that's all inclusive software so that as we continue to deploy and develop new features and functionality, you're able to take advantage of it. It's full trading credit and capacity consolidation so that you never have to buy the same terabyte twice its blade upgrade. So that as we ride the Intel curve with all
16:45
this modularity, if you want to take advantage of the next something, whatever comes after the S- 200 and 500 in the next couple of years, you're able to seamlessly upgrade to that and we do all of this non destructively in addition, flashlight is available in this very wide uh set of very flexible consumption choices. I think we're gonna hear even more about that from precaution next.
17:06
But at the end of the day when it comes to uncomplicated unstructured data storage forever. Our our whole goal here is really to ensure that flash played us is the last scale out storage platform that you'll ever need last blast definitively agreed. At least at least until at least mason can drive.
17:30
How about that? What's that? At least 10 years. At least till mason can drive. Oh wow. You're right. Mason will be driving in 10 years. Okay. So you know what is what is amazing about flashlight? S right is the ability to revolutionize some of
17:41
the most advanced workloads in the industry. Okay. So whether that's modern analytics or rapid restore and ransomware recovery, um, you know, high performance computing, B. D. A. These are really big workloads and these unstructured data environments are really what will change our world, hopefully for the better over the next over the next decade if we think
18:02
about the things we use and we think about the things that we want. Most of those are based on things are gonna come from unstructured data research. Whether that's excuse me, Unstructured data, cancer research, all kinds of, all kinds of medical breakthroughs, we're gonna see a huge change over the next two decades due to the access of, you know,
18:24
that cold data. Three years ago, I stood on the stage, many of you were probably here and there was a big blue sign behind it said no cold data, we're getting closer. That's happening and causes probably right. There's a world for tape and there's a world for flash and everything in between. That's all. Flash.
18:40
What I'm most excited about though is is A. I and you know, to to sort of co two to go along with this announcement or to celebrate this announcement. Um we're also announcing the Flash Blade Harry. S So we're partnered with our friends at NVIDIA uh to bring something special to to market. And uh we're gonna leverage the Flash Blade s the video D G X. Uh the video networking and we're gonna find a
19:05
uh fast and simple enterprise and quite frankly mid enterprise or commercial uh pre packaged and pre validated ai platform. And here to talk about our partnership is someone that I have a great relationship with, known for four years. I'm super pleased to have him here. He's a really, really smart guy. So I'd like to welcome Tony package from NVIDIA
19:26
onto the stage. Tony. So good to see you again and it's so hard not to keep saying like it's so good to see you in person on that little zoom screen, excited to be here. This is awesome in person in person. Uh so you know, we've enjoyed a great mutual partnership for several years now and I think
19:56
it's very fair to say when we introduced AI ready infrastructure almost four years ago, I mean this was really uh we were really pioneers at the time right to drive the intersection of AI and simple scale scalable integrated infrastructure. Um what are your thoughts on how things have changed and the role for aI for organizations today? You know, I think back to 2018 when we first
20:22
found ourselves in the same meeting rooms in front of the same customers over and over again, helping them solve the challenge of scaling ai infrastructure for some pretty incredible work. World leading research, uh some of the largest hyper scale type organizations out there, helping them figure out how to design complex infrastructure, solving the design complexity system. Izing it operational.
20:48
Izing and all these challenges and they essentially told our two companies, hey you guys need to get together and make this simple, make it easier to scale with predictable performance. that's manageable and out of that, you know, we gave birth to area essentially and we're incredibly excited, incredibly pleased of what our customers have been doing with that platform for the last few
21:10
years essentially everything from autonomous vehicle system development to the world's largest social media giants to recommend our systems to natural language processing. We've seen it all and what I'm really excited is that the world of AI has shifted from what used to be a lot of big science and academia, no offense to big science and academia but to pragmatic use cases that every enterprise, probably every enterprise in this room can now
21:39
power their business with. We have mutual customer in brazil City, that's a research and development arm for Samsung that's building the next generation of language models to serve the Portuguese market. Hundreds of millions of customers on their devices being able to interact in a natural way with services powered by Ai trained on Harry. So that is I mean it's exciting to have seen the evolution and whatnot.
22:04
And we've been excited to watch the evolution of the D. G. X. Platform. Everyone has just heard all about the path that we've been on with with flash blade, but we've gone from D D. X one in the initial area and you went to G. X to a 100 just recently at G. T. C. I think announced your next gen platform as
22:22
well. So we're looking forward to that. So can you, can you share a little bit more about that path you've been on and what it means for our mutual customers. You know, I was thinking earlier too, when you and matt we're talking about hundreds of billions of objects were living in a world where natural language models are going into the trillions of parameters,
22:42
right? This is to deliver super human language, understand that can power so many different aspects of a customer's business. And the reality is that today's computational problems that our customers are solving are truly data center scale. So we've had to focus intently about system izing the entire stack from the Ai frameworks to the models that we pre train to the
23:08
communication to the primitives, the drivers libraries right down to the GPU and then especially the storage that feeds that. And so we've had to think about all these layers, optimizing them and delivering an integrated stack that can power it all. And that's why I'm so excited about, you know, what we're doing now with areas, we talked a lot about efficiency in the past hour or so.
23:33
Um and I mean as you think about the elements of what makes this the right time to introduce areas to the market, you know, is that part of it And what else comes comes top of mind of why it's the right time for us to to double down on this partnership and on our joint solutions. Yeah if you look at the world, we're in right now amidst chow challenging times. It's never been more important for organizations to focus on growing the top line,
23:59
saving costs everywhere, streamlining operations. But essentially Ai is rooted in all of those things and who's the hero of A I it's I. T I T. People get to be the hero of scaling AI capability across their enterprise. So it's never been a more important time for I. T. Leaders and their teams to think about
24:22
infrastructure, to think about platform and strategy that can actually democratize access to aI development resources across an entire entire enterprise such that they can do these incredible things. So I think the timing is perfect for teams to think about area as their Ai Center of Excellence as their way to give a I development resources to all their developers.
24:47
I certainly couldn't agree more and I think this energy between our two organizations from on so many levels but especially as it relates to our roadmaps has been fantastic and thank you again so much for the collaboration on this project on the solution and for joining us here today. Yeah, thank you for having me. Thank you. You forgot exactly,
25:10
wow, thank you Tony much appreciated. Okay, um dawn of a new day. I like that. That's a good slide choice over pick that one. I'm super excited about our about our partnership with with NVIDIA. Um it's been a tremendous, tremendous partnership to this date and I can only see us expanding this in multiple, uh, multiple directions.
25:31
Look, we're entering the second decade of the transition from disk to flash. I think if you were to ask people, you know, 10 plus years ago, when when flash really began to emerge, they would say, how long is this transition gonna take? Nobody really knew, I think we don't know, you know, the last decade was about block storage.
25:50
That transition is still happening, doesn't mean that the block transition is done. There's a long way to go on the block side. But the economics have reached the point where now that's, that's beginning to happen on the unstructured data side, charlie hinted at that this morning this morning as well. Um, and for us, you know, it's that it's an opportunity for us to be a catalyst again in
26:10
this industry. We were we were we are riding a macro trend of the transition from disk to flash. But there are points in time when we as pure storage with our customers and our partners can accelerate that trend. And this is this is absolutely one of those points. So, you know, if we think about, you know, kind of what that means for all of us,
26:29
right. It means that we have to think about what are the right applications for flash and how do we get there. Okay, so Flash played as obviously building on our on our on our flash experience. You know, it brings the power of evergreen um you know, to, to modern file and object workloads is wonderful.
26:44
And you know, the time for flash plate as it is obviously now that we've seen that in the market, um you know, we're going to continue to do as well as we have with our customers scratch that we're gonna continue to over deliver uh and continue to to drive more value for our customers. But we're also going to continue to push the sustainability aspect of our business because I do believe that's an important part um for for
27:08
for us an important role for us to play as we go forward in the future. Okay. I'd be remiss if I didn't, if I didn't do this. Uh First of all a big thank you to our customers and our partners who traveled here. Thank you. Uh it's we're coming off of a two year pandemic and nobody can tell me there isn't a little bit of anxiety with traveling and sleeping in a
27:28
hotel. So I get it. Thank you for coming. Most importantly, thank you for being just wonderful partners. I love our partner community. It was such a great time at a P. F. America. America's partner forum yesterday. Great job Wendy.
27:40
Um and just get to see some of the partners over the last 24-36 hours that I haven't seen in a while and thank you for being so gracious for those of you who I didn't remember your names on your faces. You're all really kind. Nobody made me feel bad uh, to our customers. Um, you're the best. Our net promoter score is a reflection of your feedback.
28:00
As I mentioned, I'm going on my 12th year here and I can honestly say a huge part of the reason that I'm here is because of our customers. I am proud of the way that our customers use our technology, the way they talk about our technology, the way they support our company, the way they support our business and ultimately, the,
28:19
the way they believe in what they believe in by continuing to invest in our products. So thank you very much. Um, the last group of people I want to thank our actually internal pure people. Um, this Flash Blade launch is a, is a 2.5 year product um, product evolution in the making and there are several 100 engineers back at
28:43
back at headquarters and in Prague and, and soon to be in in Bangalore. Um, that contributed to this and to that team. I want to give my heartfelt thanks for everything that went into this. Um, it wasn't easy as no product launch ever is. Um, but I am just really, really appreciative and proud of the effort. The team put in, they deserve a trend of minutes around of credit for what they put in
29:05
for the flashlight launch. Okay. So uh, it's super cool. It's in the vestibule. I promise you guys can see it and read it and to go will be out there. They can show you whatever you want to show. You can pull out the DFm, you can look at the drive, it's plug and play. You can do whatever you want to do with it.
29:22
It's awesome. Uh, but you don't have to hear that from me. I thank you for your time and your attention and we're gonna roll the video and hear what some customers have to say about flash player does. My name is Mike Don, I'm an enterprise architect for the Mississippi Department of revenue. Um, Jeff Beth E Vice President I. T.
29:46
Infrastructure services for Sirius, XM satellite radio. Some of the very high bottlenecks that we had finally run up against on our flash blade. Those issues weren't there anymore. So we were able to take fast and make it faster. The workloads that we were throwing out the flash blade. Now everything just runs that much smoother
30:12
that much quicker. And again it just works being able to see the numbers and the performance that we saw and knowing that we have the ability to scale that out further faster better than we did before. Um, are all fantastic successes that we've seen in this beta. These data silos for really specific workloads. It is so adaptable and Moldable that we've been
30:41
able to slide a lot of dissimilar workloads right next to one another and they all perform well as data footprints have been growing, so has power consumption. And it's become a real sticking point for us because you know that we can only have so much electricity in our racks that we're in in some of our shared spaces. And the flash blade really does minimized consumption.
31:08
Also measured the power consumption, which is Several 100% less than what we had before. And that all maps directly over to the sustainability. That's good for the business. That's good for Sirius. XM. It's good for everybody. It's a win win win straight across the board.
  • Video
Continue Watching
We hope you found this preview valuable. To continue watching this video please provide your information below.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
CONTACT US
Meet with an Expert

Let’s talk. Book a 1:1 meeting with one of our experts to discuss your specific needs.

Questions, Comments?

Have a question or comment about Pure products or certifications?  We’re here to help.

Schedule a Demo

Schedule a live demo and see for yourself how Pure can help transform your data into powerful outcomes. 

Call us: 833-371-7873

Mediapr@purestorage.com

 

Pure Storage HQ

650 Castro St #400

Mountain View, CA 94041

800-379-7873 (general info)

info@purestorage.com

CLOSE
Your Browser Is No Longer Supported!

Older browsers often represent security risks. In order to deliver the best possible experience when using our site, please update to any of these latest browsers.