Skip to Content
22:35 Webinar

Pure//Accelerate® 2023: Building Sustainable, AI-ready Data Centers

CTO Rob Lee dives into two challenges customers are facing these days.
This webinar first aired on 15 June 2023
Click to View Transcript
00:05
Now over to Rob Lee who will speak to two of the biggest, most relevant topics in our c for our customers today. With that. Please welcome Rob Lee. Well, hello, this is great. The rooms filling in a bit more. This is, this is good stuff. We got, found some of the folks that were out a
00:33
little bit late. Uh Last night, you guys have a good time yesterday. You enjoy uh Shaq. Yeah, you guys enjoy uh our CMO Matt Bird trying to wrap with Jack and a little bit less. So um a lot of great sessions yesterday. I saw hugely packed breakout rooms, great energy in the expo floor. We got a lot more coming today.
00:52
Um You know, excited to tee up two topics today that are on top of uh really everybody's minds. Number one, uh A I uh number two, sustainability. So uh let's, let's start with A I, you know, you can't turn on the TV. You can't look at the news much less come to a tech conference these days without hearing about A I I was at the blackjack table the other night.
01:11
The dealer is talking to me about chat GP T it's everywhere. Um You know, what is clear, right? Is things are moving really fast. It's got everyone excited, everyone's trying to figure out how am I going to use this technology to change what I do? How am I going to fold this into what I do? How am I going to accelerate? How am I gonna make this,
01:26
uh you know, give me a competitive advantage out there. We've had the privilege here at pure of working with a whole variety of truly innovative bleeding edge companies in defining this space. I don't want to share a little bit of what we've seen along the way today. So let's start with A I training environments, you know,
01:45
quite simply, we are leaders in this space. We've started our journey on this space in 2017 with flash blade. We worked very closely with NVIDIA to bring out the world's first A I ready infrastructure in the air. Now, since then, we've grown considerably as Charlie mentioned yesterday, we're now supporting over 100 customers across a wide variety of A I use cases.
02:07
You know, everything from self driving cars to financial services, quantitative trading, everything in the middle. Uh Probably one of our best known customers is Meta. You guys have heard us talk about it. You've heard them talk about it. We started working with Meta in 2017 as part of research effort.
02:22
They've now partnered with us to build the world's largest research super cluster, the largest A I supercomputer in the world, right? So we've worked a lot of customers over the years. What have we learned? Um a couple of things. Number one uh performance, performance is key. Wow. OK.
02:39
You're sitting there like, yeah, yeah. Performance, of course. Um You know, A I is really a game about uh it's, it's a game of who can collect the most data, who can get the biggest data set, who can feed into GPS the fastest, who can do that over and over again. Um And it's critical to have performance there, right?
02:54
Because you've got really expensive resources waiting for you, whether those resources are GP servers that are hard to get your hands on, whether they're machine learning scientists, who you're paying a lot of money to, if those resources are waiting for data, if they're waiting to be fed, they're sitting idle, they're not doing your job.
03:10
Now, what's also important is being able to have high performance across a wide variety of different data types, right? If you look at, you know, where A I has come from, where it's going, you know, it's, you know, we're working with images, video, text, audio the whole nine yards, right? So got to have,
03:27
you know, you got to have high performance across a wide variety of different data types, access patterns, what we call multi modal, high performance. And so when you net this out, right, high performance highly parallel, it's got to be multimodal. This is exactly what we build flash blade for. Now, uh the other part of performance that's sometimes a little bit less appreciated is uh
03:47
you know, all the work that goes into managing and collecting the data before it gets fed into, you know, GPS and, and so on, you know, back in the day, there was a bit of a statistic that said, you know, in data science, about 80 85% of the work and time spent is actually in preparing the data to be analyzed. Now, I'm not sure what that looks like today with A I,
04:04
but the point is it's considerable, right? So if you think about the performance, you think about the responsiveness of all the systems that collaborate to collect data, to bring it in your back office systems, the systems to manage curate index label data, all of that stuff matters as well. All right. Uh number two flexibility, right?
04:26
Um You know, I'm not gonna stand up here and predict the future of A I. It's the things are just simply moving too quickly, but things are moving so quickly. Uh You know, the tools techniques uh data sets uh algorithms are changing day by day, right? And when things are moving this quickly, the only thing you can plan for is you don't really know what's ahead of you.
04:44
You've got to plan for flexibility, you look at where I started, you know, 56 years ago, we were celebrating computers playing chess against each other. We had hot dog versus not hot dog. We've got medical imaging, you know, these days, self driving cars, chat bots. This stuff is moving so quickly. And what this highlights is a need to invest in
05:01
a technology set infrastructure on up that's going to set you up to be able to adapt to evolve. Right? If you lock yourself into a tech stack, that's too inflexible, you're absolutely going to get left behind. Number three, uh reliability, right? Make no mistake.
05:18
These are absolutely mission critical environments. You look at the amount of money being poured into these environments. You look at the stakes, you look at the arms race that's out there, uh downtime data loss, uh it's just you, you can't have it, right? And so these are absolutely mission critical
05:33
environments. One of the biggest concerns that folks are trying to grapple with and try to figure out with large scale A I training is data privacy, data governance, right? So security controls governance, data management, all of these things become critical and then lastly efficiency, right? We know a couple of things, right?
05:52
Data scales are going to grow, you're going to have to have lots of it, you're going to have to have more GPS, GPS servers and CPS are getting more and more power hungry. Power is not growing, your availability is not growing quite as quickly, you've got to invest in efficient infrastructure. So if you net this out performance flexibility, reliability efficiency,
06:10
these probably aren't the things that you know, come to mind when you start an A I project, but they, but they really should be right. Our most successful customers in this space are beginning with the end in mind. So switching gears a little bit, um you know, not every customer out there is uh you know, going to go train their trillion parameter L L MS or build their own generative models.
06:29
But pretty much everybody is trying to figure out how they can realize the benefits of A I, right? Um Things are again evolving so quickly, you know, these models are going to cover pretty much every type of data set, you're going to have general intelligence out there. Now, the challenge is how do you connect that general intelligence?
06:46
Um How do you, how do you connect the last mile? How do you connect it to the data that's running your business? Um You know, and in some ways it's really actually not that different than uh training humans, right? Uh Think about, think about it this way you go hire a new college grad and you, you know, troll the uh you know, the best universities in the land,
07:03
you pick a, you know, you, you pick the best EMR a student, you bring them in. How effective an employee are they day one, right? They're pretty smart but they don't know anything, they don't know anything about your business. You've got to train them, you've got on ramp them, training computers,
07:15
not that different, right? So the key to the key, the key to getting good results here, the key to making use of this technology is connecting the A I powered General Intelligence, those algorithms, those applications to your data. So data is key. Um you know, there was actually a paper, a memo that was released leaked. I don't know it was posted out of Google
07:38
recently. It's pretty interesting, it's called we have no moat and there's a lot of stuff in this, in this memo. But one of the conclusions, one of the points of the memo was, you know, in the future, the secret sauce isn't, you know, all the neural networks and the training and the GP us,
07:52
the secret sauce is going to be in the data. The winner is going to be the folks that you have the high quality data domain specific that are deeply attached to the use case to what is being trying to what's trying to be achieved. And so we know that data is important, you need lots of it, but it also can't be cold, right? The whole idea of collecting lots of data for A
08:13
I is you've got to actually be able to access that data. So it's pretty clear that this is a space where you just simply, you know, have no opportunity to have cold data. It also can't be siloed, it's got to be accessible, right? And so, you know, again, I keep telling you, I can't tell you,
08:28
you know, the exact future of how A I is going to play out. But there are a couple of first principles that are pretty clear, right? Number one, no call data disk simply just has no place here and the data has got to be warm. It's got to be on flash, it's got to be accessible and it's got to be accessible based on things like security and
08:47
policy and compliance and governance, not infrastructure limitations, right? You've got to get yourself away from silos, you've got to implement more of a cloud operating model. All right. So let's switch gears. I'm getting the red blinky light here. Um Sustainability.
09:01
Second most common topic on on top of customers minds Charlie put up some numbers yesterday, you know, 1 to 2% of data center, 1 to 2% of global energy spent in data centers. A bunch of it's in storage. I've seen estimates that in the next decade that might even double so big, big numbers, lots of energy. Our ability to produce that energy is not scaling to put this in perspective.
09:26
What do we have 113 terawatt hours to put this in perspective by the end of the decade, the amount of energy consumed by data centers is about enough to power every single home in California. Uh Coz mentioned Doc Brown earlier I was doing some math behind stage. This is about the equivalent of sending 100 or 100,000 or so Deloreans back to the future.
09:48
So energy is hard to get it costs. A lot of companies are taking this seriously. Half of publicly listed companies in the US have set long term energy reduction goals. So many of your employers governments are getting involved. You know, we're seeing data center build out expansion, power draw bans moratoriums from all around the world. Ireland pore, even parts of the US,
10:13
right? And this is reflective of a couple of things, but mainly it's that energy is hard to get to, it's hard to produce, right? We just don't have enough of it. You know, in the sustainability conversation, sometimes you know, there's a lot of elements of it. You you hear people talking about carbon
10:28
credits, planting trees, renewables, you know, I'm a simple guy. I think you basically have three strategies here, right? You can either reduce the demand, right? Use less energy, you can improve the supply, go to renewables, you can mitigate the impacts like plant some trees. Now don't get me wrong.
10:44
All of these are good. Planting trees is great. Do that. Renewables are great here at pure, we use entirely or almost entirely renewable energies, but the biggest impact that you can make, right, the biggest way to break through the limitations is just simply to use less energy. And that's really what we're doing for you.
11:03
So the net of this is, you know, energy space efficiency, these are real limitations. You know, you simply in a lot of places can't get more of it. If it's not already part of your buying decision and technology selection, it absolutely has to be, don't get me wrong. This is as much of a limiter as speeds and feeds as features. You absolutely have to be certain with this in
11:26
mind. Now, I've talked a lot about power, we talk about power a lot, but the other element of this is also reducing footprint and e waste, right? If you can, you know, deploy denser solutions, more effective solutions from a space perspective, that's less space you got to pay for in the data center, less materials that have to be produced,
11:45
sourced, shipped, it's less cost and labor to go and manage that. And that's less waste at the end of the day, less landfill that we're generating. All right. So how do we get there? Um You know, from a storage perspective, it's about a couple of things. Cos and Amy talked about direct flash showed you the density,
12:00
that's a huge huge lever, right? We talked about 100 and 50 x improvement over the last 10 years. That's 150 X. Um the power efficiency that we've driven data reduction, right. We've pioneered ded Dulic compression. Those have just gotten better and better and better over the years. That helps too lastly evergreen.
12:19
Well, ok. How does evergreen help? Well, evergreen, the promises we deliver in evergreen, the ability to grow and scale without taking disruption, um eliminates one of the most wasteful elements of technology buying, which is over buying, right? You don't have to overbuy on day one. You know that you can expand, you know, you're not gonna get left behind and we supercharge
12:39
that with Evergreen one, right? With Evergreen one, we'll go and work with you, we'll go and work with you uh to deploy the gear just as you need. We'll go and work with you specifically on meeting those energy. Uh S L A s. All right. So how much can we really save here? Uh Charlie put up some numbers 2 to 5 X better
12:54
than all flash, uh better than 10 X compared to the systems that we're typically replacing. Um But let's let's talk about what customers are telling us, right? I was meeting with a customer recently. Um They measured the impacts that they were seeing after deploying peer gear, it came back and said we're now consuming on the storage side of the house.
13:12
85% less energy in our data centers. We're producing 97% less e waste. Forget the number here, 96% less space on the systems that they had before. So my point is it's absolutely significant. Right? And the best part of this is we're just getting started. You heard coz you heard Amy, they were talking about where we're headed with our density road
13:37
map. We've got the opportunity in the years ahead to further improve these benefits by an order of 345 X and beyond. All right. So I'm way over my time. Sorry about that. Just to recap. We hit a lot of topics today A I, right. If you're planning on building,
13:52
start with the end in mind, performance, flexibility, reliability efficiency. If you're planning on using, think about how you're managing data. No more coal, data disk has no place here. Implement a cloud operating model, right? And then all around. If you're not thinking about space power, energy efficiency as part of your technology
14:11
selection and buying decisions, you absolutely must be. All right. So next, I'm excited to welcome to the stage, a customer who's made an incredible impact in reducing his company's environmental footprint in a moment. I'm going to bring Ajit Sharma who is the business Optimization manager for Virgin Media two to the stage.
14:29
But first, let's cue the video, the wind through my tree tonight. She leads me to me with the she's taken.
15:21
Why? Walk when you can ride, introducing the fastest wi-fi guarantee of any major provider. Please help me welcome Margaret to the stage. Thank you. Uh All right. A before we get started, a couple of things I noticed, uh, they didn't give us the shack size couch, so I'm a little bit,
15:44
uh, a little bit upset about that. Uh, number two, uh, you know, those socks are awesome. Can we get the camera guy, uh, coming, uh, close up on the socks? Look at that. You got the pure logo in there? I, I'm only pointing this out because we're a little bit competitive here and I just want to
15:58
make sure that everyone knows that some of us prepare our customers a little bit better with a swag. Thank you, really appreciate it. Thank you, Rob. Excellent. Well, welcome, welcome to the stage. A thank you for spending time with us today. Uh Could you maybe just start by telling us a little bit about the Virgin Media Two story?
16:14
Yeah, sure. So happy to be here everyone and good morning to everyone in Vegas. So we merged together in about two years ago with a company called Virgin Media and 02. So we're the Uk's biggest Telco with approximately 46 million customers in the UK. So we offer quad play. So that's anywhere from TV,
16:30
through to your mobile phone, through to fixed line and broadband. So if you think about it, every two out of three UK customers is one of our customers. So that's, that's, that's quite big. But where we've been really successful has been the huge demand in customer insights and some of the A I S and the analytics that you spoke about earlier.
16:49
Excellent. So, you know, we clearly very topical matters. You know, you guys have been focused on analytics, how do you blend your business? But you guys have done a lot in sustainability as well, right? Really focused on driving efficiency. Can you maybe share a little bit about, you know, what you've done in the past and kind of
17:04
what's changed over the years? Sure. So we've been a pure customer since 2016 and we embarked on a journey whereby we realized that sustainability and going forward, the climate plays a big impact. And we were proud to choose pure storage due to a its technical delivery b its ease of use. And then the third one was CS G and E S G.
17:25
It was unbelievable. We reduced our data center footprint by 98% 98% 98%. So it's like 50 X, come on, 50 X guys huge. And I think the key one for us was our reduction in power. We saved 96 per cent. So we're in the high nineties.
17:47
And if you look at where the market's going right now, in terms of energy costs and everything else to see a huge reduction in our power consumption cost is unbelievable. Plus it helps the environment. So we could think of a better thing to do than to buy fuel. Wow. So those are some amazing numbers you know,
18:02
I, I'm curious because often in this conversation, uh you know, I talk to customers and, and they're kind of working through, how do I balance, uh what's best for the business versus what's best for, you know, sustainability and efficiency. How do you guys think through that? So, it's, it's changed and developed quite a lot in the past few years.
18:19
So when we used to run RFP Tenders, sustainability, maybe used to create maybe one or 2% of our overall waiting. It's now nearly 25 to 30%. And that's where pure does really, really well, your credentials and your credibility really helps us maintain that. So for us, it's a huge change. It's so important.
18:38
If we don't make steps now, we're going to hurt the future. So now is the right time to be doing. It couldn't agree with you more. You know, I talked a little bit about, you know, folding these concerns into everyday kind of uh technology decisions, buying decisions. Um you know, uh is this a top down mandate? Are people all throughout the company uh
18:56
thinking this way, does this, you know, how, how do you create a culture with the company uh of, of driving to these outcomes? So, one of our key strategic actions and objectives is we're going to become carbon zero by 2040 and we started that journey already by replacing all our fleet of vans that currently use gas and diesel by now using all electric and these are all sustainable energy. So we started to make that step,
19:19
I guess what's really been important is from top down, there's an investment decision, money needs to be made available to drive sustainability. Without that money, sustainability is really, really difficult because as you can imagine changing from A to B you have to go through a journey that not only just means a technology change but also a change in culture change. Culture is really important.
19:39
It's got to be driven from the top to the bottom and make sure that everyone's involved getting everyone on that journey start with the finance guy, right? You've always got to go with the finance guy first and that's always the difficult one. Yeah. Yeah. Well, you know, on the other hand, you outlined very significant savings,
19:53
50 X reduction in space. You know, what do you say? 96% 20 X reduction in power that must drive tangible cost savings. Is there a virtuous cycle there to kind of put those dollars back to work into investing in other sustainability goals? Absolutely. So in the last year alone, we got rid of over 48,000 tons of E waste. That's equivalent of nearly 16 adult rhinos.
20:22
So just trying to put that into perspective. So it's quite huge. And what we're seeing is the money that's being saved from e waste and opportunities to reduce our costs that's being refunded into other parts of the businesses. So as you can imagine, we've seen Four G, we've seen five G. Well, guess what guys, six G is going to be out soon as well.
20:39
So we need to reinvest our money in the right places and this certainly helps us on that journey. Got you. So you, you talked about, um you know, E S G and, and kind of the considerations of uh around that area, driving a lot of uh decisions and investment. Where have you guys seen the biggest outcomes and benefits?
20:57
Energy savings pay uh space, uh you know, sourcing e waste like across the entire spectrum of sustainability. Where have you seen the biggest benefits? So for us, I think it's the case around as you well know, the UK is a much more of a smaller country in terms of size by us able to reuse and use sustainability. We don't have to go and build more data centers.
21:17
So we're helping, we're helping the UK and the community to ensure that we're not using more space that's required. So we're saving millions of pounds by not building new data centers. We are focusing and staying where we are now but reducing our footprint and we'll be able to do that. Thanks to Peel. Awesome. Awesome. Well, so in the last couple of minutes,
21:36
uh you know, I'm curious what's next uh you know, for the future uh in Virgin Media, you know, you talked to six G, whoops, cat cats out of the bag. Uh What else is uh what else is next, what else is in the future? So, as you can imagine, you know, it's an exciting journey. We've just merged um merged two big companies into one.
21:51
Uh We've got an exciting five year road map strategy and most of that involves around ai automation. And above all, how do we continue to reduce our costs? And that's where we're looking at A I significantly. And I think so with the help of pure storage, we can certainly look at different ways of
22:08
doing things. And one of the things I'm really excited about which I've learned in the last few days is your, is your E series. I think so. That's really going to help us reduce our footprint even more. And above all, I think they make a good commercial decision for us. Excellent. Excellent.
22:21
Well, thank you for your time today. Um Everybody, a round, a round of applause for a. Thank you so much. Thank you so much, so much. Thank you. Thank you.
  • Video
  • Pure//Accelerate

Rob Lee

CTO, Pure Storage

07/2024
Pure Storage FlashArray//X | Data Sheet
FlashArray//X provides unified block and file storage with enterprise performance, reliability, and availability to power your critical business services.
Data Sheet
5 pages
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.
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.