00:02
Hello and welcome to the April 12th, 2023 Coffee Break. My name is Andrew Miller, your host. As always the principal technologist here, excited today as always so many great guests uh to be joined by Justin Emerson, Flash Blade Technology evangelist. We're gonna be walking through like it says there, the disc dinosaurs never saw it coming. You know,
00:22
having a little bit of fun, hopefully never jumping the shark or being too snarky. But you know, having a little bit of fun with the meteoric impact, a flash blade e and inside my head, I'm thinking like, man, I'm saying meteor meteoric meteoric, like what's the right way to say that Justin, you want to help me out there? Did I get it?
00:36
Right? Although generally like you use meteoric in the context of a meteoric rise, which if you think about it doesn't make sense because Meteors don't rise, they don't, they don't ascend. They usually it's quite the opposite. Maybe we should have thought of that before the title. I don't know, but it felt right at the time.
00:55
So diving in a little bit of opening housekeeping as always, this is a series Thank you for joining us for what if I'm keeping track? Right? Is the 28th coffee break. Uh We're a little bit past two years, you know, happy second anniversary. Always like to put this up just because because of the solution focus these my opinion.
01:12
But hearing from other folks too, even yesterday at a partner event in Atlanta, these have aged better over time. So both the educational side, the career advice, the actual technology, the architecture, feel free to go back and look at some of these. You can also find them at pure stores dot com slash events. There's a whole events center which lists not
01:29
just the coffee break series but a bunch of other great series that are different styles, tech talk flash crew. You can as well find previous links there too. If you don't use the bit link that I had earlier, now, I would not be a good pure citizen if I didn't make sure to highlight that in just about two months, I'm pretty sure I'll maybe it's the next time I'll see you in person.
01:47
Justin at Accelerate. We're gonna be at the brand new resorts, World Las Vegas. So without further ado, there's an amazing 42 2nd video. Thanks for the thanks to the Accelerate team. Let's let's roll.
02:42
Got away for the last couple. Uh Oh no. Now I'm missing the instrument. Shame of the xylophone RBA. Yeah. Thank you. It's like, oh the last one as always, I do know that folks. So, so please make sure you join us for accelerate if you can.
02:57
If not, there's gonna be a good bit of the content online. Also do realize that uh hey, there's a whole coffee card thing to this $10. You receive those. After this, the the 1st 1000 participants, there's some folks in the exclusions list, you are super special to us but you know that we can't send you coffee cards for the obvious reasons.
03:14
Wrapping up the housekeeping. I always like to do a little bit of even just not, not a full introduction because if you've attended a coffee break, you don't need to hear about Andrew every time. But in this case, when we start to talk about hard drives and disk, like actually when you and I were chatting, Justin like we both had some help desk
03:29
background and I remember taking hard drives for some professors at the university I worked at and freezing them usually for a day or two give or take uh because they, they'd stuck up inside section kind of stuff. The, the heads are incredibly close to the platter and that sometimes would be enough to get the drive to spin again that you could actually get the data off in their last quiz or papers or whatever else they needed for their
03:52
classes, you know, kind of thing. So that, that's more just like Andrew from a hard drive perspective. But Justin, do you mind? Uh do you mind briefly introducing yourself? Sure, thanks. So, uh Justin Emerson, uh I've been with pure since 2020. Before that I worked with uh uh in the channel
04:08
community and reseller partners. Um uh Down there is my uh my lovely wife and son. Um And uh prior to, you know, my professional career I will work at or I worked. But prior to that, I actually went as a student to uh University of California Irvine. Um I majored in Japanese which surprises a lot of people.
04:31
Uh And then that screenshot down there in the lower left was I used to uh play uh dance dance revolution competitively. They definitely have that on me. Um We had it for the wii and never did very much with it. So, you know, look at that and, and just to be upfront for everyone, we're not gonna explore dance,
04:48
dance revolution theme very much like that. You can contact Justin separately. He's got his email there, you know. But hey, so yeah, yeah, but it's cool. Next month. Uh we'll be joined by John Brett. If you remember about two years ago, we actually had Paul Ferraro on to talk about
05:02
Evergreen, well, pure as a service. Now, Evergreen one, we're actually gonna be talking a good bit about storage as a service, any kind of subscription economy, what pure has done there recently in this space. And actually John he actually used to work with customers to help write their rfps or the requests from big vendors around various as a service offerings.
05:20
So we'll walk through that a good bit. But this month it's all about disk dinosaurs, flash la. And of course, Justin, we will be doing a little bit of the normal agenda uh like, you know, you know, not, not too much a agenda. First part is often like to do a little bit of career exploration. I appreciate Justin being,
05:38
being willing to kind of go there from a personal standpoint of some of his career and the inflection points there, you know, kind of being just you couldn't, you can't stop fixing things sometimes, then we'll actually do a continuation of something that Justin's collaborated a lot with Brian Gold on around the Better Science Series, some more of that content that links into, you know, just between us,
05:57
our, our super secret plan. If anyone knows from a uh a AAA from Electric automaker standpoint, uh There's one of those out there, you can go Google on it. Let's talk in some about flashlight D and then focusing on hopefully what matters most to you, which is not just the products and the technology, but where would you use it and how and why please make sure to put questions into
06:17
the chat throughout. We've got some great folks on uh Olivia Emily, Andy Bish that are here to help with Q and A and chat and we'll have some Q and A at the end as well. If they don't get to your questions, maybe even if they do, we may read them back out. We'll be done at around 45 minutes past the hour.
06:31
Let's go ahead though and kick off if you don't mind, Olivia, we'll kick off the first poll here before we get started. So this is, to be honest, um, I appreciate that we have the latitude for a pierce damn boy to have a little bit more fun with the po questions sometimes. So, what's your favorite dinosaur?
06:49
Um I was thinking when we were writing this Justin of the scene in one of the Jurassic Park where the kid's watching Barney while his uncle, I don't know, is getting chased through, you know. So, uh I, I would definitely answer. I prefer cats, but I do love dinosaurs too. You're the preferred. Yeah, cool.
07:06
So we'll just let, let that run for a minute, you know, and, and you can ponder you, you, you can give deep thought and ponder to your answer. I wanted to start out with really a little bit of your, your personal history, Justin. So, I mean, you've been with pure for a little while, but there's been a good bit before that.
07:22
Do you mind even kind of starting back a little bit with? Um, there was partner side, it was even back at, I think university of California, Irvine. I kind of narrating a little bit of your, your history. So, uh so, yeah, so as I mentioned, I, I uh I went to college in uh in Southern California at,
07:37
at UC Irvine. Um While I was there, I became a student worker in um what was called administrative computing services at the time, which was the it department for all of the internal um sort of business function of the university. So I didn't deal with faculty, I didn't deal on the academic side. Um And uh but rather I worked more in like,
07:59
you know, hr and accounting and all of those kinds of, you know, departments that every kind of company or organization has to have. Um But uh you know what, um what I remember most about that time was uh I was part of the, the, the team that did the first uh VM Ware virtual infrastructure 3.0 deployment in the UC system. I think Santa Barbara did a 2.5 install before
08:24
us, but we were the first to show off, you know, H A and the motion and all that really cool stuff and really, really revolutionary. And uh I remember thinking, wow, this is the, the biggest thing since sliced bread, it's going to be in every data center in like three or four years and, and I was wrong, it was two.
08:42
And so I ended up um going to a partner to go do like VM Ware installs. Um And because I was like, that was a really fun problem to fix. Um But now I want to go like, fix it at other places because there's more interesting challenges and that was one of the cool things about the, the partner side is you got to touch so many other, you know,
09:03
environments and customers and stuff, which if you're working at a, at a, at a single customer, you kind of have your environment and that's, and that's sort of the, the world that you live in. Um So I started out doing professional services and then, you know, started seeing things that like, well, I feel like there's a better way to do this
09:21
problem is being someone in the field. So I'm like, started moving into presales and, and then I started saying, well, there's so much more around this VM ware environment than just the VM ware part. There's the networking and the storage. And so I started going and trying to do those things too. And um so I, you know, I've always go and look
09:42
at problems and say, how can I, how can I fix those things which sometimes drives people a little bit nuts. Um And, and sometimes you learn after being somewhere for a long time that something is not actually a problem or even though you view it as a problem, it's that way for a reason that may be a good reason or it may be a bad reason. And if it's a bad reason, once you finally realize that that's sort of when you say,
10:07
ok, I can either choose to ignore it or I can go somewhere else because I find this problem is intractable. So there's a little bit of we all vote with their feet eventually. And, and especially of the comment of that when I started was actually a gentleman who had been in it for 20 years, but he is really good, humble guy. He's my boss and he would regularly help with
10:26
explanations of like, I see something like what in the world is that? Like, I'm right out of college and be like, well, we did it for these reasons at the time and we want to come back and revisit it, but that was the best we could do then and there's a lot of rational thought. So, uh, so keep going. So, so at VM Ware at, at a partner doing VM Ware things and,
10:41
and I think if I remember right, you were, you ran across pure storage? Actually a little bit. Hands on. Yeah, actually, um, we, the, the partner I was working at, at the time we were one of the first ones to, uh, sign up with Pure, I remember I got to be on stage at, I think it was the pure sales kickoff in like
11:00
2015 or 2016. The one that was at the Marriott in Santa Clara. Although I think a bunch of them were, um, and I got to except a little trophy on behalf of my company and it was, uh it was great, but I remember working with some of the very few first puritans um down in Socal who are, who are still here. Interestingly, at least uh a few of them.
11:19
Um And so, uh yeah, I've been working in sort of the pure orbit now for um close to 10 years. There's even the comment you made too earlier when we were chatting about just that, especially on the partner side. I mean, what the one of the keys to success and it resonates for me, given similar background is to always think of the customer as an extension of you and their
11:40
success is, is your success. I'll, I'll let you keep going there and, and then a little bit of how you came over to peer. Well, you know, when you're something, something that people don't often realize is when you're at a partner, the, the thing that you have to offer your customers is your expertise and your um trustworthiness
12:02
because you, you do services and those things are based upon your expertise and how much you're trusted. But generally speaking, the stuff that you have to sell is stuff that somebody could buy from anywhere else. Um because you don't, you don't fundamentally like make anything like a vendor does So if somebody doesn't like you anymore, they can just go get that from somebody else.
12:23
So literally, the most important thing is your reputation and your um your expertise to your customer. And so the way that you facilitate those things, the way that you foster and grow those things is by taking a, a really uh you must take a customer first mindset where you say, how do I do everything on behalf of my customer? That's one of the things I like about pure.
12:49
I'd never worked at a vendor before I came to pure. Um because I feel like we have very much a similar philosophy behind that. I've, I've worked with, I remember being on a call with salespeople from some other company who will, will not be named and they said we should position this product here. And I said, why? That doesn't make sense?
13:07
They said because that's the one I get paid on. And I found that disgusting. And I said, I never wanna be in that kind of a company or situation. And so far, Peer has never, never been like that. We've always been very customer first. It it helps. It's possible it's made possible by having good
13:23
products that are designed for customer center but not, not perfect ones. But there's, there's a whole product and capability and that enables that mindset when you bring that mindset maybe last last, but not least I know I'm, I'm letting start out a little bit. Um you came to appear, but it was partly because you actually saw a preview.
13:39
I think of the, one of the, the, the origin of what we're here to talk about today at a product level once we get there. Yeah, so, so um uh because we had an amazing channel team when at the partner that I was at, got to when I was up in the Bay Area for something else. Uh I was invited to Pure Headquarters and I got to meet uh John Hayes, one of our founders um and he showed a prototype of what eventually
14:04
became flash bla um and I uh thought it was just again, I had that same feeling about it that I had with, you know, VM Ware V I three back in 2005, 2006, which was, this is just amazing technology and I really, really wanted to, to, you know, be a part of it. And so, you know,
14:26
when I eventually did come to Pure, I came in support of Flash blade just because I thought, I think it's such a, a cool product and say nothing ill of flash array. Flash ray is also an amazing product. But um from a, from an architectural standpoint, the stuff that flash blade does is just I think so cool, so pretty. I think there, there was some comment about
14:48
that. Uh not everything needs fixing and you and your wife have conversations about that periodically, but maybe we'll let that one go and keep going, but I had to put it in. So, uh, actually, so let, let's actually, I'll close up the first poll here and we will launch the second one. So let me end that and share it back just because, you know,
15:05
hey, we're gonna have fun with it. So, for anyone who's wondering, hey, uh, t rex edges out velociraptor, We have a healthy population of folks that just prefer cats a little bit in chat. Uh Bar Barney actually has a respectable showing, although Barney's, you know, maybe aging a little bit. Um Olivia, if you don't mind, actually go ahead and launch pole two there and this now starts
15:25
to go into a little bit of what we're gonna be talking about today. So, um, we're curious, this is actually your brain child ESTIM because we're getting into thinking about flash and it being pervasive. Do you do? I mean, for me, uh I'll just say I only, the only hard drives in my house at this point are for backups,
15:43
I think maybe for you, it's a media server. Is that right? Basically. And, and it's not for lack of, of, uh, of trying to, to replace them. What would, uh, you know, I, uh co coz our, our other co-founder, uh, liked to point out that even for everything that used to be on spinny disc in his house in
16:02
his house, he's since replaced them, even if it has been at, at the time, at great cost. But I guess when you're the founder of a successful company you can afford that, um, it feels like an internal consistency thing too for. Yeah, exactly. I mean, you got it and, and he, um, he did a demonstration at our sales kickoff and he used
16:18
the hard drive and he said I bought this used on ebay so that I will not give any of my money to the hard maker. But I, I asked, I, I thought this would be a fun poll question because sometimes people, um don't have, don't think about how much computing has changed because of flash. Um, there, you know, if, if you looked several
16:43
years ago, you know, you might have a spinning disc in your, in your DVR, you might have it in your game console, you might have it in your desktop computer, all that kind of stuff. Right. You might have hard drives in all of these different things. All of those things are now either flash based or in the cloud.
17:04
Right? I, I, it's hard to think of a consumer electronics device that still has a hard drive and that's actually something that's really different between now and, you know, and 10 years ago when, when you're a little bit more than 10 years ago now. But when, you know, a decade ago when we were first getting into into the market in a big way
17:24
as pure storage. Um You know, we we were like on that um that trajectory of flash. And if you look at now, why is flash which we'll talk about later? Why is flash getting cheaper? It's because it's in everything. And people say what's the next thing after flash?
17:41
I say, I don't know. But whatever the next thing after flash is, it's gonna need to be in everything in order to achieve the economies of scale. That flash has because flash hasn't gotten inexpensive because of data centers. It's gotten inexpensive because of phones and tablets and TV. S and cars and all that stuff.
18:01
So we wanted to continue a little bit partly based on the poll from January and a little bit more of the better science content because there's a whole wealth of stuff there. We actually promised that at the end. So if you want to go back and listen to January, Brian Gold walk through a good bit of this. But actually I'll, I'll let you, I feel like you were seguing into this thought a little bit
18:17
just on how data has changed. And, and I'll just let, let you keep going from here if that's all right. Yeah. So I think if we had to encapsulate what better science means to us, the, the idea is, is that it's, it's how we approach solving problems. And I, I think the, the, the way that I try and encapsulate as in as short a time as possible.
18:37
What the real problems are for. We'll narrow the scope here to just say enterprise storage. Um is that the nature of data has really changed. Um We talk about the explosion of data. We talk about data is the new oil. But when we think back to what data was many years ago, it was created at human speed.
19:00
So you had individuals, you know, a transaction would happen because somebody would buy something or some record would be created because of some interaction that people have. But basically new data got created because of um human activity. But that's really not the case anymore. Now we generate so much data and most of that data is generated because um
19:26
of things that systems do on their own sensors, uh internet of things, devices, the phones in your pocket, the app, like all of these things are constantly writing new data. And so when we talk about that data has exploded, the reason that we say there's so much more data now is because we're writing so much more of it.
19:47
It seems kind of obvious. But what that means is that we've gone from a environment where most data was where, where most of the the IO was reads and not writes to now a new world where there are so many more rights. There's a lot of really cool like computer science stuff and how you do data um data placement and, and build data structures to orient it more towards update driven um
20:14
systems but only file systems. Yeah. Yeah. But if you think about that in the context of the change in media, which is what we've been talking about, right? Not even talking about like floppies like behind me. But you know, the hard drive uh is about seven years old as a as a concept. This is AAA lovely picture of one of the,
20:33
the very early ones. Um And the big problem with hard drives was always seek time, you know, like you said, you used to put them in the, in the, in the freezer. And that's because there was a physical head that has to move back and forth around, right? And so everything in computer science terms was
20:50
organized around avoiding disc seats, which happened when you read. So if you have data that was gonna be read all the time and seeking was slow and you needed to seek in order to read stuff, you had to build everything around that constraint because disc seeks were, were the enemy that is not the case anymore. We do not live in a world where the majority of enterprise systems like this.
21:16
I'll say systems, not bits because we'll come back to that. Um But they use media that is solid state and where rights are the constraint. It's not that they don't write extremely well, but it's that there is a finite endurance and that means that there's like this conflicting trend. We're writing so much more data and we're
21:39
writing it to media that has a finite endurance. Um and that endurance is getting worse every generation of man as we go from SLC to MLC to TLC to QLC to maybe someday PLC or something like that, right? This is the core dilemma that every modern storage platform has to deal with. If you're using um flash as your media and you're living in a world of
22:04
huge data ingest and data um and data update rates. How do you square these two things because they're, they're sort of in opposition to each other. And our point is that using old systems with design for old media and then using them with new media, it is uh put a technical term on it dumb.
22:30
Um If you are using old data structures that are designed around avoiding disc seeks, they usually do that by implementing lots of different things that write more in order to read less. Uh this is where things like copy on, right? And all the other kinds of technologies come from, not just to guarantee, you know, data consistency and resiliency,
22:54
but also to allow for navigating these data structures in as few disc seats as possible. But then we're running that on top of flash masquerading as a hard drive because that's what a solid state disc is. This is what Brian talked about on the last coffee break in January is the entire idea of a solid state disc is to take a bunch of flash and put this um this translation layer, this flash translation
23:22
layer in between it and the system in order to hide the fact that it's not a hard drive. And so it appears to the rest of the world the same way a hard drive would, which is logical block addresses, which is just a way to map sectors and tracks from a hard drive. And so yeah, this, this causes so much inefficiency. It causes so much extraneous right?
23:46
Amplification, which is in other words, if I'm gonna overwrite the same four K sector over and over and over again, hard drives work like that, but flash doesn't work like that flash has to do things at a block level, it has to do erase blocks and all this other kind of stuff. And I'm not gonna rehash all of that stuff you can check out, you know, the the second did but that putting an old,
24:10
old data system, an old data structure on top of new media pretending to be the old thing. Um like you said, you know, when you were working and you know, say why were things like this? It's like, well, at the time that was the thing to do and if you look at this, you can see, OK, I understand how we got to this point because it was inertia in all these
24:33
things. But at some point, you have to be the chronic fixer and you have to step back and say, is there a way to fix this problem? And that's actually what pure storage was founded to do. The whole idea behind our company was we should build systems designed for flash from the get go and not build systems that are retrofitted or systems that
24:59
have to accommodate both things. You need to build systems that solve these problems. That because these problems are only going to get worse over time, even back in the early days of SLC and MLC flash, we could see where the, where things were going and we could see that if you don't solve these problems today,
25:18
you will have to eventually solve these problems later and it will be much more painful to do it later when you've built up 10 years worth of scaffolding on top of what is effectively an unstable foundation. So that's what purity, that's what direct flash. That's what all of our design philosophy is about. This is why we call it better science because
25:40
we think that this is a better approach to solving the problems of modern data storage. Um And I would imagine that if you look at this logically, I don't think you could really disagree that in a vacuum, this is clearly the better way to approach it. I'm gonna go back into a little bit of high school and taking Latin uh Tabula Rasa, right, another term for clean sheet. But the idea of that sometimes you have to go
26:06
back to first principles. It's not easy. And a lot of times companies that have been doing this for 20 or 30 years, they're dealing with innovators dilemma type stuff, but pure we, that's where we started 10 years ago and been working on it. So, OK. And, and we didn't have an existing business
26:21
model that we needed to defend. Um Similarly to as we'll talk about with, you know, electric vehicles, if you're building internal combustion engine vehicles and you're making lots of profit doing so and there's this really disruptive technology, you're not really incentivized to, to rock the boat too much. So that's where disruptors can come in and,
26:41
and really change, change the market. So I said I did a Latin illusion, but I think we've got a almost an Egyptian illusion with the pyramid though. So bring us home in this section. Yeah. So, so the when, when we talked about uh the pyramid, I know we did all this direct flash stuff on on the last time when,
26:59
when Brian was here. Um But there's a great article I highly recommend I say great. I wrote it. So I'm a little biased, but there's an article I wrote, which is uh all about uh it's called Better Science Volume Two. And it's all about what the pyramid is, and the pyramid is our internal um log structured
27:16
merged tree based database system. It's our key value store that we use. It's designed specifically for flash. I did a podcast actually recently with um Fong Wang who is a pure employee number three. Although that's debatable is if you want to know why you can listen to the podcast. Um But um what's really amazing is when we started the company,
27:40
we started the company and said, we need to build a system that's designed for flash. And so we need to do that all the way from the ground up, including how we lay out data and how we build our, our system. And so um the, the goals that were set out back then um are things that we've pretty much achieve. We built a system that went from, we built,
27:58
we built a structure that's fit in systems that are tens of terabytes to now tens of petabytes. And so we've gone, you know, a thousandfold scale designed specifically for, for all fashion. Um We've had customers who've been on a single array through many generations, but one contiguous identity of an array now for, for more than 10 years.
28:21
So people have really been with us on, on this journey. And the combination of these two things is what we mean when we talk about building a system designed for the ground up with hardware and software, um you want to map the things that your system does as efficiently as you can onto the hardware that does it because that's how you do things like use less energy. That's how you do things like use the,
28:49
the product that you have, use the components that you have the most effectively. Um And this is where I think software defined systems um while they are great from a flexibility standpoint, they have to sort of build everything to the lowest common denominator. Whereas what we do is by engineering, both ends of the of the spectrum, we get the maximum amount of efficiency out of
29:13
everything, sort of in between. And so it allows us to map the portions of our data structure, our log structured, merge tree directly onto the hardware components that we do, that we design to run them. Um And that's, I think as far as I know, unique amongst the storage industry. So with that, we're gonna pause because this is actually gonna segue almost directly into
29:39
section three, but I'll go ahead and end and share the poll. This was still having a little bit of fun, but it was actually meant to help highlight and thank you Justin. And I just feel like every time it's positive. Thanks. That was awesome. Um There are a good number of folks that have a lot of hard drives or only for backups.
29:57
So, you know, actually the majority, it's only for backups. That's interesting. We're 50 uh plus the ones who say no, it's a hard drive that puts us at a majority so we can interpret any poll the way we want to, I guess. No, no insults to the folks that do have lots of hard drives. We, we're not saying you look better and max,
30:11
but no, that, that was just having fun because hopefully we, we have a, we have a little more license to do that, but it does mean that the majority of people out there aside from, you know, backups on a on a shelf kind of thing, don't, don't have hard drives in most of their devices. So Olivia, if you don't mind my launching poll number three and uh we'll go ahead in there.
30:31
OK. So this is where we get, we get a little more practical someone saying, hey, like the question was a little bit ridiculous with Beta X. Yeah, we're having fun with it. But in this case, how much disk is? Not in your house, but in your data center and we try to think about the best way to kind of categorize this
30:44
for you based on quantities of based on size, not based on numbers or quantities or rack units. You can put other comments in the chat, please as well as what do you use disk for today? Like why do you, why do you still have it there? There, there has to be a purpose for anything in the data center, it has to be justified and the business case and use case.
31:01
But here we actually want to continue on into and, and maybe we'll see if we can make, we'll, we'll, we'll pull this section a little shorter if we can. Justin too thinking about it is flash blade, but there's a reason we're gonna call it, you know, just between us, our super secret flash blade plan, get to that in one second because we actually want to start with continuing to explore that
31:20
really the hard drive theme. And what is it used for in the data center today? So back to you Justin. Yeah, so despite the fact that, you know, we've been very successful with, you know, introducing flash into so many parts of of data centers.
31:36
Now this includes data that's stored in, say, for example, public hyper scale cloud providers and stuff. Um but 90% of unstructured data still today really does live on spinning disc and that's from a capacity standpoint, right? Which makes sense, the majority of the capacity is, is still living on, on on slower media and we use much faster stuff for the the kind of
32:00
thing that that needs that performance, right? And so we think and, and the analysts are the ones telling us this, we think that data is going to grow something like tenfold by the end of the decade. Um And if that's going to happen, I mean, you look at how much spinning disk is out there, how much power and all those other things that uses, that's,
32:23
that's really oops, that's really not a uh a, a sustainable approach. Um Because disc inherently um has certain characteristics, there's a physical platter and that platter spins at a certain number of revolutions per minute and there's heads and all these other like actual physical constructs. Um And so if you are looking to grow your data footprint by 10 X and that and that your
32:53
drives are not growing in size by 10 X in that time, that same time period. Because if you have a 20 terabyte spinning disc today, that would mean having a 200 terabyte spinning disc by 2030. And that's I don't think anybody's talking about that, right? So, and not only that because you have physical
33:14
platters that are spinning and motors and all these other kind of things, this is gonna use a lot of energy and it uses all that energy even at Idol as well. Um And because there is so much complexity in building huge systems out of all of these fundamentally unreliable mechanical components, there's all this complexity in, in building huge disk based systems.
33:39
And you know, I can imagine that most people would rather not run spinning disk given all other things being equal, they would say, well, I'd much rather have all flash for everything. But, but why don't people have all flash for everything? Why do, why do I still have hard drives in my media server?
34:04
There wasn't another option, historically, even getting some comments in the chat. Just basically money. I think Brian there wrote the dollar symbol like 10 times that, that feels about right, maybe 20 times if I had, if I had the option of a 10 terabyte hard drive and a 10 terabyte SSD and they were the same cost. I'm gonna take the SSD every time.
34:21
And so what people really want is flash, but at the price of disc and what would be even better is then getting all those total cost of ownership uh benefits along with that, the lower energy costs, the better simplicity, the higher reliability, et cetera. And, and so this is where our uh I, I'll turn it back to you and you can talk about what uh
34:43
our super secret plan was. So, so, so for those who aren't aware, I I'm, I'm not gonna give free shoutouts to companies that make way more money than that. I mean that, that, that are very well established in the industry, but there is an electric car company out there where if you Google, they actually talk about for years back,
34:56
their super secret plan to world domination and all the other things and going to the moon. And no, but actually now I'm conflating different companies that the same person might own, you know, but there is, it's out there and it's a fun way to think about it. So we decided to kind of follow this a little bit because ironically, even before some of that was published it maps to what pure does.
35:13
So let's, let's start on the, I don't know, maybe the electric car side to illustrate it if you don't mind. Right. So, so if you think about what the super secret plan was to um grow the viability and grow the market share of electric vehicles, which I think the news today was that um we're we're looking to um move the United States just, just talk about us. We here, we here in the US,
35:38
move the United States to almost two thirds of electric vehicles by 2032. Super exciting, but you don't get there until you reach economies of scale. And that's the, that's the idea. So you start out with something that's premium, that's very high end, that may be very expensive. But you use that to prove out the technology to make it better to make it cheaper.
36:00
You start selling more things with more use cases that starts moving down the stack in terms of where it's viable. And now you look at electric vehicles and, and yeah, we're still only above 5% of new vehicles like of light duty vehicles were uh electric in, in 2022. Um But that's a huge improvement over zero. And if you look at how the adoption curve is going, it's getting,
36:27
it's, it's increasing and it's accelerating, which means that yes, we're, we've only gotten to 5% but that's the hardest 5%. Believe it or not until you get to the very end because it's this like sigmoid curve thing. Right. The chasm crossing the chasm stuff for no one who's gone and read that book, read it. And as,
36:45
and as that gets better it actually puts pressure on the economies of scale of the thing that you're replacing. Um And that's where we see flash going. That's where we see flash has gone. Flash started out very expensive in the data center was only useful for the highest tier things in the consumer electronic space.
37:06
It was only useful in the areas where functionally hard drives would not work if I wanted to make an uh you know, a music playing device that was thin enough, I couldn't put a hard drive in it, even one of those little micro drives, right? That uh I, I needed something even smaller, even more embedded or I needed something more durable, something that could handle vibrations
37:29
or I needed something that could allow me to make a laptop that you could, that was so thin, you could slice bread with it. Um And, and so over time flash has gotten into more and more stuff and, and, and not just in the data center but everywhere flashes in everything. Um It's in your refrigerator. If you get one of those smart fridges,
37:50
right? So um as the economies of scale get better and better, again, it puts pressure on the economies of the, of the the status quo. Um And you can look at the, the recent um you know uh how, how the the disk makers have been doing. Um This is where flash blade E comes in is that this is for us
38:14
the first time that we think there is a all flash based platform um which is going to be more environmentally friendly, more economical, easier to manage and last fundamentally longer than a disc based alternative. And so this is the culmination of our super secret plan. Um is we're, we're finally at the part where we're ready to go take out the last tier
38:43
of disc. Um I talked to possible maybe 10 years ago. Exactly. I talked to Pete Kirkpatrick, who's one of our VP of, of engineering, of hardware engineering. And he said, you know, this is every, every time we've developed a new platform or launched a new platform or introduced a new drive type and all these things,
39:01
it's always been about taking down the next tier of hard drives. Um There is no tier after this, there, there is no tier beyond this other than that, that is spinning media, I suppose you could call, you know, there's linear media which will not get it. But um you know, there, there is a certain point beyond which there,
39:21
there is no longer any um any more, more, more, more lands to conquer, right? Anything spinning at 5400 rotations per minute or 7200 rotations. But I mean, the, the, the key here and I wanna make sure just to highlight for folks is this is flash, we're going after a hard drive focus and the way that we're doing it is we're making,
39:42
I think for the first time, some pretty bold statements about price points publicly. That's right. So I mean, we have always been um you know, when we're, when we're talking about what pure provides to customers, we always think that we provide the most value in in what we do for for the the products and the services and the solutions that that we provide.
40:04
But in this market, as I mentioned, the only the the business differentiator that spinning disk has in this category is not any of those sort of value add things that's purely economical. And so that's why with this product, we are talking about the economics sort of right up front because that is the thing that will move the needle, that is the decision point that people are making,
40:29
we can talk about the performance of it, we can talk about, you know, the the simplicity et cetera, but the actual decision criteria that push people towards disc in this market is purely economical. And so we have to talk in the terms and be OK with that because that is again, the thing that, that drives um you know, the, the decision criteria for environments
40:54
like these multi petabyte scale unstructured data repositories, things where data goes and it tends to sit there and not a lot happens to it. Again, it's mostly rights not reads. Um So this is kind of our, our stake in the ground to say, you know, yes, everybody can go out there and talk about how they're going to be cheaper than disc, this is where we're sort of putting our money where our
41:20
mouth is and saying this is where this is kind of where the starting point is. Um where our flag in the sand is. I think I may, I may summarize a couple of spots here if that's all right. All right, Justin. Um So, I mean, we, we are continuing to see crazy increases from a density standpoint especially and we've even commented some
41:40
publicly on this actually about, you know, our direct flash modules and how much we see that capacity increasing even up. Do you see it on the chart there? Like 300 terabytes, it's public. This isn't just a, a space thing. Um It also has dramatic environmental impact. So that cost is a cost per gigabyte,
41:56
but it's also power savings and e waste and operational costs. There's less things to fail there and have to go into data centers and fix and even the space from a rack unit standpoint. All of this is built on top of, we couldn't have done it without all the work we've done on both the flash layer as well as flash blade software over the years.
42:14
So how does the replication work? How does the object work? How does NFS work? It all works fundamentally the same way. And if you're curious for depth there, uh Brian gold, actually, it wasn't meant to be a multipart. Brian Gold. Shout out, you know, about a year and a half ago, two years ago,
42:29
we went over the like the core of flash blade architecture. If you don't mind, Justin before we kind of get to use cases, I'll, I'll turn it back to you to kind of talk about. We have a, we have a couple of different sizes here, but we're trying to keep it pretty simple and I summarize a little bit there. So feel free to put in anything that I left out
42:44
while I was in summary mode. Yeah, for sure. So um you, you know, we uh as I mentioned, like you'd need like a 200 terabyte hard drive by the end of the decade to meet those things. We're looking at, you know, 300 terabyte direct flash modules before the end of the decade. In fact, our our founder coz on the launch webinar for this said he sees one petabyte
43:05
modules by the end of the decade. So really exciting stuff now. Um And what that means for today is that we're launching flash blade e in three different sizes, small, medium and large, small, not being actually all that small when you think about it, it's, it's about uh 2.6 petabytes usable without any data reduction. So it's gonna be better than that. So it's about four petabytes raw,
43:28
six petabytes, raw and eight petabytes raw. Um And these numbers will go up over time. We'll have more sizes, we'll have um you know, larger drives later than this year. Um So all of these numbers are, are, are going to go up from here. And that's because in these environments, people are talking about this amount of data. Um You know, the and, and when, when data is going to grow by tenfold by the end of the
43:51
decade, you know, you're gonna need all of that all of that space. So that's our uh our, our uh our objective here is to get to get bigger faster um than um all the other solutions on the market. This is part of an overall portfolio. If you're, if you're asking about performance, uh flash blady is,
44:12
is not the conversation we want to have. We probably wanna collect the flash blade S series that is performance focused in various ways. And this is also available via Evergreen once this can, this is still part of our full on consumption based model if you want it to be and it really rounds out our entire portfolio, our portfolio. You've got flash ray X and XL and C.
44:34
We've got Flash Blade S and E think we're gonna wrap up with use cases. But any uh any final thoughts there and just kind of the flash blady overview, Justin. Yeah. Yeah, I saw somebody in chat asking about like file serving speed. Like if speed is really important to you, we already have that product that we launched that last year with the,
44:51
with the Flash Blade S and the launch of Flash Blade S is that architecture was that enabled us to now deliver Flash Blade e less than 12 months later. Um Because of all of that work we've done up front with the um with, with the, with the engineering on that. So and can't emphasize that enough. It's so much if, if it feels like rapid product release, it's because there's so so much work
45:16
done on the foundational pieces last but not least we had fun with the title, not competing the Dinosaurs, but which niche um are we thinking of even witch witch and ice cream sandwiches or sandwiches, all this stuff? It it and Whit Witch, right. Yeah. Yeah. So, so you know, there are all these use cases where we've been positioning flash blade s for sort of high performance tiers.
45:39
Um But every single one of these areas has large chunks of data, the larger parts of folks data estates that aren't living on flash today that are living on spinning disc. And so if you're thinking about where does Flash Blade e play? It plays where the on the home turf of spinning discs.
45:58
So like it, it, you know, we, we see it as it's gonna out compete disc in that evolutionary niche. Um And, and we think that over time, it means that we're though that that area is just gonna erode for um for the legacy uh storage um solutions. I think sometimes of a lot of these applications they have need for hot tiers,
46:22
whether it's, you know, data protection, the most recent seven or 14 days, you might need to restore it really quickly in the event of a ransomware attack. We're talking even analytics. A Splunk has a warm tier concept. So this actually can be mapped across and then a lot of them have long term retention of data. You need to keep longer. It can't be locked away in a,
46:39
in a vault on a tape, but it needs to actually be better than hard disk or it's been on disk and you didn't, it's just the only spot for it. So that's kind of the classic model. But then there's nuance for each one of these where we see it playing out. If you, if you think of another use case, please put it in the chat. We we would love to hear from you around that any any other thoughts?
46:58
Sometimes our coolest use cases are ones that our customers bring to us, which is, which is really, really fun. That's, that's where safe mode started and a lot of the rent more stuff. So that's, that's the story I tell anonymized because it, it should be kind of thing.
47:11
So I, I don't know, Justin, I feel like we should like take a try to announcer voice style of where an all flash data center is no longer a dream, but a reality, I don't know that we're the right people to do that. You can try running down the dream of a of a all flash data center. Let's actually, I will um we'll close up. Well, actually let's go ahead and,
47:30
and um we'll share poll number four and then we'll wrap up here a little because there was a fun one here. So I'm gonna end the poll and I'll share this one back just for folks because we were curious, you know how much disk is in your data center. Um This is pretty evenly distributed, although there's 6% that's all flash. It's kind of fun to see. Hopefully everyone is seeing this.
47:51
And then what do you use for disc today? I feel like somehow the um I don't think the answer is sort of automatically, I think the first three that we thought of for the top answers just in um enterprise imaging data protection and analytics. Um And there's a drop off after that and I'll, and I'll just point out that, you know, uh there are um folks with um and not that I think it's almost the majority,
48:15
but maybe not. Um I have uh an, an, an environment where they need the kinds of capacities that the flashlight e is already delivering. So, Olivia, if you don't mind launching Poll number four and then we'll wrap it up with a drawing and some other pieces. So, um but this is frankly what, what we do want to hear from you.
48:33
Um It is actually that same list, you know, because what use cases are interesting for you, that might not be the same answer as what you're using it for today. You know, it's like I've been there, done that, but I have these new needs. Um And even just in this space, what does drive when we decide to have a little bit of fun with green as in money, as an environment as in how much experience do you need?
48:53
I was your put your can you put your greenest tech on, on managing the system? Like fresh out of college? Like I was coming out of the help desk and you too actually way back in the day as was I, so we went through a lot here and even to be honest, we chose a little bit to spend a little bit more time,
49:13
even some of the, the career back on your side, Justin and even the better science, we had a great launch webinar last month around flashlight. Even it goes even deeper into some of the pieces. But we want to make sure to hit the highlights and the focus and the use cases. Thank you, Justin for being a great guest. As I was as I was sure you would be,
49:30
you know, I'm always sure, but then it always works out. This is live for those folks that are wondering, this is truly live. We don't actually, you know, record it and pretend to be live, you know, kind of thing. But for anyone who's stuck around for the drawing, thank you, uh drawing or the Q and A.
49:44
Um let me make sure here. So the winner of an ember mug value $130 the kind you can control with your phone is Alexander W from Michigan. Thank you. And before we get to Q and A, just because I didn't, I, I didn't do it. I just like, oh, I'm in drawing mode in my head.
50:02
Any final thoughts here, Justin before we uh jump into Q and A, I, I wanna give you last word before we formally move into Q and A. Well, I just, I'm uh I'm thanks for having me. I'm super excited about where we're going. Um You know, one of the reasons I'm here at pure storage is I really do think that we're changing the game, we're changing the industry and that it's something that I can do that uh
50:22
Every day I can go to work and I could be proud of the what we do and what we stand for because it's not just, we're not just interested in the all flash data center because it's cool even though it is, we're interested in it because it's a better way to do things. Because if you look at the energy savings that you get from moving to all flash, it's, it's really gonna make a huge dent. Something like 2% of power in the entire world
50:45
is used up by data centers and a quarter of that is storage. So if you can make a dent in half a percent of global energy usage, that's doesn't seem like a lot. But, you know, we need everybody rowing in the same direction to really get to where we need to be. So thanks for having me.
51:02
Oh, thank you. So with that, uh we have a little bit of time left. Uh And that's always the goal. Usually we aim for around 45 minutes sometimes as we cheat back and forth, then I summarize stuff. But there was so much good stuff you're saying. So I don't want to cut it off hard earlier, Justin.
51:15
Um But because we intentionally time for Q and A. So there are um some questions and actually, um I think Jason, uh I always appreciate Jason folks asking for the sign up link for the next coffee break. Hey, that feels good. You know, it was a good. So actually I do want to make sure to highlight that next month.
51:31
We'll be back with John Brett, uh talking about storage as a service, evaluating the industry overall at a kind of an analyst level than what peer does there plans useless planning and dispensable. Um I'll assume that actually Olivia or Emily can toss that link in the chat. Let's maybe go over to the Q and A because most of these um Bish and Andy have been great about
51:52
typing answers for. But actually some of these, I think I may uh we may kind of revisit here a little bit Justin or maybe that'll be our opportunity to like, oh I wanted to put this thing earlier and I forgot or Andrew cut me off, you know. So uh there there was, go ahead, go ahead. Sorry. No, no, go ahead.
52:08
Um And so this is actually a fun one because we do get it commonly and we'd rather just embrace it and take it on a question from Theodore. Is there an HSM or tier that's needed or anything like that? Where when I think of, you know, you see a, you see a slide like this will kind of go back and there's different price points and performance. So how do we think about tier?
52:31
Well, um I think that if you looked at the, at the use case slide, what one of the things is that all of the use cases that we're going after in this sort of initial period are ones where storage management or, or um you know, the the data management moving between different classes of storage is handled by for the most part, the applications, which is something that you pointed out,
52:54
right? And so while we don't have anything to announce today, with regards to, you know, HSM or, or things like that, one of the things that I will point out is that I'm as, as I'm sure someone previously on coffee break has talked about, we do have this vision around fusion and the idea for fusion really is to make it so that
53:13
you don't have to think about tears that you don't have to think about any of these things. And we're not there yet today, especially not with, with flashlight, not saying we are. But if you think about where we want to go, that's really where we want to go is we want to have this be part of the larger story around automatic storage and storage policies and, and all that kind of stuff.
53:34
And so while, you know, we, we're targeting these applications that kind of have hierarchical storage management um integrated into their workflows, um that won't be the case forever. I think that I can embrace as there is a great question from Stacey about uh F Bs and FBE into single cluster. I sometimes even answer that as well on a
53:53
Flasher a side for what it's worth. Um The pure does we do? Obviously believe in tears? We haven't believed in tier, especially kind of in Interra tier that's classically sometimes on the block side. And a lot of that actually, to me is a little bit of a philosophical thing. It's this embracing that you want to know when
54:09
you do any read or write a deterministic response from a performance standpoint, not knowing or not, not being able to know. Where is that data living under the covers? And how might it respond this time based on an algorithm that's kind of a little bit opaque, right? So obviously we embrace different tiers. There's a philosophical approach to Inra tier that, that I'm actually very comfortable with
54:29
having designed and built a lot of these systems over the years. Yeah, I think that when you look at at any kind of sort of hybrid system, whether it's using um tier or cashing or storing of metadata on separate you know, media and things like that, when you start thinking about that, every single algorithm, every single clever optimization et cetera will always
54:54
find one edge case where it breaks. And if that happens to be your edge case, whether that's today or someday in the future, it's a rude awakening when you, when you hit that. So we try and design our systems in a, in a consistent uniform way. So that that doesn't happen. Great question from Ken. I think I know who he is.
55:15
I was trying to be careful about last names in case people aren't public listening, but I think it's Ken with one of our great partners. Um and actually was asking and this is just, I mean, it's a good question if design is based on flash players, there's commonality from the software standpoint, the software layer uh with the major difference being some of the underpinnings,
55:30
the flash et cetera, because it's all the horsepower to support things like rapid recovery, et cetera. I feel like I've got to pull out the, it depends and you might kind of laver what it depends on and, and, and I'll say that if you, if you look at that use case slide, right? We still talk about flash blade S as our
55:47
solution for rapid restore and ransomware recovery. The four Rs um where flash blade E is positioned is in that long, longer term data repository. Now, would it would it work for your particular needs? It depends like you said, but from a positioning standpoint, rapid recovery, rapid restore ran recovery is still in the realm of
56:10
flashlight s. Um and it's not necessarily about like a horsepower thing per se, but really, it's what are we tuning the system for? And flash E is designed for maximum efficiency, whether that's power efficiency, whether that's space efficiency, whether that's cost efficiency, those are, you know, just happens to have the E in the name too.
56:32
Go, go figure. Uh it's like somebody thought about that whereas flash blade s really is about that speed. And so if, if you're optimizing for different things, you end up with different permutations of the same system. And that's why we have two different products. That's why we have, you know, flash blade s and flash blade e both useful in the same use cases.
56:54
But in different ways. The other piece I think of there sometimes now this is I'm channeling a bit of a conversation I have with Andy Stone. He was on a previous coffee break around ransomware. Sometimes what he'll talk about is that there needs to be that rapid recovery tier. You're bringing back the most recent within the
57:08
last 37 days to restore maybe 50% of your environment in mass. But you still will need your instant response team to be able to go back and look at 30 60 90 100 and 80 days back with pretty good performance to be able to see how the Attackers got in what they got access to. That's not just logging, it might even be older backups to the system.
57:26
So there's even a both end approach there uh from a ransomware and rapid recovery standpoint. And when you're doing forensic analysis of log data or analytics or all that stuff that's, you know, way in the past when you're doing queries on things, those are, you know, things that are metadata read intensive and those are things where a seeking hard drive head is gonna be real painful, I think with that we
57:52
are right at time. Uh It's 59 minutes past the hour for whatever time zone you're in. I want to do one last comment and plug for uh joining us next month uh with John Brett storage plans. Useless. Planning indispensable. That is an illusion to the uh Dwight D Eisenhower quote for anyone who likes that.
58:11
Storage is a service and you uh next month on May 10th, Justin. Thank you so much. I'll uh I'll let you have the last words to bring us home. Awesome. Hey, it was great being here. I'm so uh so privileged to be part of uh this um this great organization and what we do and I
58:28
hope uh all of our customers, all of our prospects, all of our partners um are, are looking forward to taking the rest of this journey with us towards y'all flash data center. Thank you so much. Awesome. Thank you all. Hope you have a pure storage this month's coffee break. Have a great month. See you next month.
