00:00
Hello and welcome to this month's coffee break, Thursday, April 14th, uh time zone, depending on where you are. Uh My name is Andrew Miller, excited to be joined by JD Wallace today. It only took me a year to get you back. JD, something like that. Well, you know, I mean, we've been on so many cool projects together,
00:18
the Pure Report Unplug and a bunch of other webinars. So, you know, it's been a while and I'm really excited to be back. But uh you know, we're still old friends. We do this all the time. We have a lot of fun, don't we? We, we talk outside of Holy public venues that are recorded with lots of people.
00:31
Otherwise it would be awkward when we did this kind of thing. So, you know, as always, um this is meant to be a coffee break or you can relax. Hey, there's some free coffee cards involved, but it's also a time where we explore topics that we think you'll find interesting or we seek to make them interesting. First half is always educating you on areas. We're going to do a little bit of a technical
00:50
foundation on pure today and then move into some pieces around um, recent Flasher updates. You know, how, how pure helps in this space more specifically, hopefully, feels more like a podcast than a webinar. We're gonna get a little bit of opening. I don't know if it's all housekeeping out of the way,
01:05
but for anyone who's looking at the title, the Power 42 anybody um wanna put, you can put it into the chat. We thought about doing a poll but it was like, hey, you know, we don't need to do that. Uh The Power 42. I'm betting some people know that this is a joke referring to Hitchhiker's Guide to the Galaxy, right?
01:21
And I, I think JD, this is where even you, you and I felt like we were such a good generational boundaries, you know, such a good movie, right? That's why everybody knows it from. Oh II, I might have read it when I was a kid in this kind of nice leather bound edition. So we were thinking that they are the folks who had listened to this from a movie standpoint or
01:38
watched it, she watched it or read it. And if you know this, this is where there's the, there's the story of the uh computer that goes away to figure out the answer to life's ultimate question, the universe and everything. And the answer is 42 right? In pure parlance. Actually, we have something that we call the 15
01:54
architectural decisions we're gonna go there and it was like random number that might not have any meaning. And, you know, sometimes when you're coming with titles late at night, it felt like a fun thing to do. So, playing off a little bit of some previous um kind of it. Joke history. Although I don't know if I'm dating myself and
02:07
yes to uh for Michael in the, in the chat. It is. Uh yes, the the it is really, what is the question as always in your Coffee break series. This is a series. We are now up to number 16. I think if I'm counting that right.
02:22
Um You were back in number two JD back in February of last year. Ever so cool. It's been a fun run man and it looks like you even have a new logo and everything. I mean, this is really kind of become a fun production. So again, honor to be back on the, on the program.
02:37
Thank you Emily. Thank you Adrian and thank you for your marketing for, we continue to iterate and as, as you all keep coming back, we continue to make it a little more official as you know, there's coffee cards involved. Uh when you sign up, you'll see that by no later than tomorrow from Emily.
02:51
If it goes to a personal email. Hm Can't resend. There's some exclusions, folks that we can't send this to. And then of course, if you stay for the end. There is a drawing for an ember coffee mug, which actually I happen to have one with me here today. So you might see me taking a sip out of this periodically.
03:06
Now, full disclosure, I actually chose not to turn on the heating function because that felt like one more thing that I didn't want to have to try and keep track of while we're doing a live, a live presentation. So you, you don't want these self own of spilling hot coffee all over yourself in the middle of 844 participants. We're up to now try to keep it relaxed. My brain's running kind of fast.
03:26
I was like, I don't think I need one more thing to be a tipping point. So as well, next month we'll be joined again actually by Jack Ho, we're going to revisit. Um Some of what we've been hearing is we talk to customers about their cloud journey um because it's challenging for everybody and there's a lot of different options and
03:42
obviously a lot of things that pure does. So we're talking about the first half of actually what we're hearing when we talk with customers about their cloud journeys and then even some updates on Equinox Metal, what I talked to Jack about last month as uh last year as you know, um your host, Andrew Miller. In this case, I, I don't want to make my intro about, you know,
03:59
the stuff that you see here, we use the same slide every time. But actually something that I found very interesting and this is even starting on the customer side. It's once I dug down deeply enough into products and architectures, sometimes that foundational understanding of the architecture helped me figure out how I could use it in different ways or maybe in ways that I
04:16
shouldn't use it for or know when I was pulling it further apart from its, its design center fuel, we're gonna come back to that. So for me, this is actually even on the customer partner side, seeking to kind of deeply understand technology, partly because I enjoyed it. But then also found that it had, you know, a good impact over time.
04:32
JD. Do you mind uh with kangaroos? That's just a happy fun picture. I'm a human. And uh if you didn't get to see a kangaroo today, now you have and in the US that usually makes people smile, you know, they're not pests over here. So, you know, and, and, and that's what I love about all the conversations that you and I have
04:48
together. We've done enough of them now that I think a lot of that personality has come out. We've learned a lot about each other and we've shared that with some of the folks that uh that, that come and join us on these things and, you know, we are real, we're real human beings, real lives outside of pear story. Hey, we love it here. We have a lot of fun.
05:01
We do a lot of other cool stuff too. And so I, you know, I like, I love, I love learning about that aspect. So I think that's really fun that you put that in there. This was at the Caversham Wildlife Park in Perth, Australia. So there's your extra little, you know, fun fact.
05:15
You mind reintroducing yourself JD. Yeah, I am JD Wallace. I am a principal technologist. Um You know what Andrew and I are actually on the same team, which is one of the reasons we get to work uh so much together. Um I tend to cover a lot of stuff on the west coast of the United uh United States.
05:31
Um Andrew, of course over there on the east Coast. Um But yeah, uh just about to celebrate my two year anniversary here at Pure Stor. It's been a heck of a ride first time. Uh actually being at headquarters today, you may have noticed a little bit different background for me. I'm actually at pure storage in mountain view. So much fun to be here and,
05:51
and you know, you, you talk about those personal, those personal aspects of who we are, those personal connections. You know, it's so much fun to be here to actually talk to some of the engineers that are building the really cool products that we get to talk about on these webinars to get the meet on. And, you know, the there, there's so many little aspects of the things that,
06:08
that they work on, that may not be what we, what we really think about from a customer enablement, more, more the back end kind of stuff. And to just hear some of the engineers get to talk about the cool stuff that they're building that you may never actually see. It. Just, it's, it's so much fun and it, it puts that personal touch that you talked about to all the people
06:25
that are here uh kind of working to, to do what we do for our customers so much. It's fun. But yeah, I'm, I'm, I'm JD uh two years of pure gosh, long background and all kinds of fun storage stuff. Maybe some of those stories will come into play at some time even did a lot of work in data protection like like our friend Andrew here. Um Yeah,
06:41
there's some other fun geeky things that I like to do. Andrew and I, we talked about this last time. Both super into LEGO. I'm actually taking a little PTO next week gonna go to a Lego land myself. So I'm really looking forward to that trip. Very cool. Like how I paused right as you were taking a
06:56
drink, Andrew. Excellent timing. So let's dive in like it said power 42 and 15, fulfilling the potential peers, technical foundation, what we wanted to do today is, is potentially try to mix it up, make it a little bit different each time and you know, keep some continuity but mix it up.
07:14
And I almost kind of gave myself permission today along with JD of in the first half, what we wanted to do is actually look back at some of the core pure architecture. Some of this links into uh the session that we had with Sam Marini about five months ago, we're talking about legacy storage shackles. Then we're gonna dive into what we call the power 15. There were actually 15 architectural decisions
07:35
that go back to 2009 underneath pure. When I when I first came to pure, it was like this was fascinating to read. We're gonna play some of these out. This is a little bit of feeding our technical souls, but there's also a lot of practical impact on these. So this is frankly kind of the first half of educational stuff about storage architectures and things you have to think about when you're
07:52
doing that, then we'll transition a little bit into looking at flash array, recent updates. So a lot of the 15 architectural decisions while they relate primarily to flash array, they don't exclude flash blade. But you know the history of appear here started with flash ray, right? We all know this hopefully.
08:08
So we'll look at some uh recent uh flash ray updates around hardware says all about the bezels there. I think JD. You were able to find a beer too. You want to pick your favorite one there and putting you on the spot? Oh, my, my favorite. Ok. Well, I, here, this one, this one still has
08:23
that new bezel smell. We'll get into this a little bit more. Here's our flash ac, I love this kind of gun metal gray that we've used on this one. This is a lot of fun. But, yeah, I've got a here at headquarters, I got a couple of toys today. Actually, it actually lights up. But I, I think you've got a,
08:36
oh, cheap jokes. You've got an electric personality but not quite enough to light up that bazzle. So, you know, and then last, uh, this is actually referring back to the Mark Andreesen quote, you know, about software eating the world. So, you know, software world eating, we'll flip that around and talk about some recent updates from a
08:51
software standpoint with flash ray. So, and as always, please, uh, we've got folks, uh, Jeff Emily Sean Zain to help us with the Q and A. If you have questions, please don't hesitate to put them in the Q and A. If you want them truly answered, put them in the Q and A to go and see them. The chat usually goes by really fast,
09:06
feel free to have fun in the chat, right? But we're not necessarily watching for the Q and A. So I'm trying to keep an eye on it kind of thing. So let's dive in first. As you may remember if you listened back in or not listened, you know, if you joined us, you know, because everyone's here,
09:18
we're all, we're all busy. But if you joined us back in last November, you may remember uh with Sam, we did a session called Enterprise Breaking free from storage, hardware shackles, you know, the history of Evergreen for pure when we had some fun with, you know, the idea of, you know, breaking through the limits of legacy storage architecture. And some of that was the concept of that,
09:38
you know, that there can be hopefully enterprise storage for everybody and that, you know, the value of the size of your organization or the size of your workload shouldn't dictate what features you get access to. III I know JD when you've, you've been at a couple different storage places here when you see this and even think about, you know, different storage platforms and architectures
09:59
and you know, SMB to mid range to enterprise, you know, over to you for a second, you know, and I think what we've done here, it's, it's not to be taken lightly because look at the end of the day, pure storage is a business, right? We exist to make money and how do you do that? It's really easy to say. OK, well, we built something super cool active
10:18
dr that's something we're gonna talk about later, brand new feature we brought in really fabulous capabilities there. It's so easy to think. Well, hey, let's, there's, there's a ton of value, let's monetize that. Let's let's go ahead and ask for a little bit more revenue to, to compensate for all of the development energy that went into that product.
10:33
But, you know, I think along the way we made it really difficult but I think really strategic decision that's that, that's really shown kind of the culture and the values of pure when we said, you know what? No, that some of these features, we don't want them to be exclusively for only those customers that can afford, afford the quote unquote high end storage. We really want that enterprise experience for
10:53
all of our customers. And you know, II I think regardless of what your business is doing, so you have a small, you know, small storefront all the way up to, you know, multinational corporation, you still need some of that value. You still need protection from ransomware. You still need the ability to move your data
11:13
where it can be most effectively used. And so, you know, breaking down the walls of what is enterprise and, and making sure that all those businesses have access to those features. You know, I think that that's really at the core here and, and I love that we made that decision because not everybody did.
11:25
It was the right long term decision, not necessarily easy in the short term, right? And sometimes you're like, do I have the balance in the short term in the long term in that we, and, and now there's still a little bit of a replay of some slides that you may have seen before. We talked about the idea of statelessness and
11:38
how stateless actually enables these pieces around, you know, not having storage, refreshes and migrations and outages and, you know, stateless is evergreen. Statelessness was actually a core architectural principle that became before even the Evergreen uh marketing term. If you will, it's, it's not a not badmouthing
11:54
it, right. It's just what we call the whole program. And last, but not least in there, this is a slide that you may have seen a couple of different times. What I wanna call out is over here where it says with pure 97% of a raise purchased six or more years ago, still in production. And this is not in the sense of where I
12:10
sometimes was in my previous life where it's like, yeah, I still got it in my data center six more years. But if you pushed it off the dock with a short drop, I wouldn't mind, you know, kind of thing. It's actually still in place has been updated and modernized. So we take some of that and where Sam and I didn't go, you know, five months ago or so is where we want to go today,
12:28
which is actually looking at how is that possible and not just being true on a slide and not just sounding good, but actually under the covers being true from a capability standpoint, II, I know JD you've thought and this is where we're going to section number two already, you know, so the power 15, these were architectural decisions and before I even, you know, put up the, the, the 15 decisions, there's this concept of MVP and minimum viable
12:54
product, which is a sometimes a double edged sword. That's, yeah, absolutely. You know, we're, we're here in Silicon Valley right now and you know, so many software startups are out there and they're thinking, hey, I, I've just got to get the MVP, the minimal viable product out there to, you know, to start making an impact,
13:09
start collecting data and that I can build from that. And that, and that's a really common software development concept. We hear about it a lot. But when you think about something as complex and I don't mean complex from a user perspective, we'll talk about that but, but storage is hard, right? Making sure that you uh honor the,
13:25
the required integrity of that data set. That that's a hard problem to solve. And if you don't put thought into how you build that architecture, your MVP is gonna set a path for you that may not be flexible enough to deliver on some of the value you want deliver 10 years out like we are now. And so II I, what I like about the the architectural decisions here is like,
13:48
yes, we started with a particular set of capabilities and we've grown tremendously over the years since then. But a lot of them go back to these core, fundamental decisions that we made, that established that path that we would go on that, that set us up for success, right? The the other piece I think about there. And by the way,
14:07
I see two folks who've raised your hands, just a heads up. Um please go feel free to put stuff in the chat or the Q and A so good for you on raising your hand. But at the same time, we we in this format enough folks can't actually pull in, but glad you raised your hand. The other piece that I think of here is uh and I actually wrote a blog post about this a while
14:22
back around the idea of design centers. It was, is my way to science, think about, you know, minimum viable product. But like any given product, when you look at it, you figure out when it was designed maybe 10, 15 years ago, what was it designed to originally solve? What kind of problems? What were the technologies that were available,
14:38
the open source libraries, this or that, you know, all the stuff that went into it? And then over time, how far is the product flexed from its design center. And that actually may mean how kind of complex it is, how hard it is to use for a certain thing. But it was designed for 10 years ago, isn't really what people want to do today, but it's still around doing very different things.
14:55
In the first use case, you think about that idea and you layer that into when you as a customer data center architect. When you're making strategic long term choices about what you're gonna put in your data center. It's not just a concept that I used. Really, I I at that point, I didn't even know the term MVP. I thought it was a design center.
15:11
So we wanted to do today is we're gonna cherry pick a couple of these. We're about to know the next next 10 minutes or so, maybe 15 and we're gonna group some of these together. You see these, you know, architecture matters, 15 architectural decisions. They truly go back to 2009 with pure.
15:25
Sometimes some of these are ones where uh some, sometimes there's competitive, you know, hot shots being taken and some of the intentionality just being real. We sometimes see that with the controller architecture, you know, front end, active, active and back, end active standby rather than read all of these at you. We wanted to do is actually take these and we're gonna do it in kind of five chunks and
15:45
highlight them first. So starting off with the concept of simplicity. And by the way, this is pure stuff, I realize that. But the goal here is that we're gonna be more about what's the fundamental architecture and that you could take some of this and apply it as you're looking at other storage products in your data center, other products in general, these are design principles that had to be
16:04
thought about even before pure when pure was in stealth mode. So number one simplicity, I love the quote here, there's a quote from uh the co-founder of the Peer formulated by kicks, who was your first product management and marketing about the idea that when you look at the very beginning of Peer all flash from the beginning and the concept that, you know, no one was going to take a chance on a startup unless
16:26
there was a really compelling reason. And so that reason back then was performance often, it was focused on VD I and databases and VM Ware, etcetera. But you wouldn't try a startup unless there was a crazy size something in this case performance. But that's not a long term competitive differentiator. The long term differentiator that would keep
16:43
people around is simplicity because like you mentioned JD, this is actually complex stuff storage inside is complex and often outside it's pretty complex too. So we're gonna actually borrow and you'll notice some of these slides are copyrighted from 567 years ago. That's intentional. It's a little bit of like,
17:01
we're not just making this up for today's webinar. There's so much stuff from a customer use standpoint that we, that you don't have to think about as a pure customer, but the array still has to figure it out. This is even that idea of, maybe we've talked about simplicity and intuitiveness too, I think.
17:17
Yeah. And, and what I love about this slide is sometimes it's what you don't do. Right. Because you know what we're not out here to do is we're not out here to look and see what the industry built and just built the next version of it. I mean, there's certainly an evolution of some core concepts, but we are trying to be really innovative here.
17:34
And so, you know, so there, there's a couple of examples of things that we just said, you know what that, that, that doesn't make sense and tomorrow's architecture. And so uh we're gonna say no to that. And that's, again, that's a hard decision to make, but sometimes it enables us to go invest those limited resources into the things that really matter that really uh move the needle for our
17:53
customers. And so, you know, yeah. Uh Andrew, um you know, to your point, there's a lot of complexity here, but we don't, we don't want to expose that to the customers. You shouldn't have to worry about all these little knobs and adjustments and, you know, hey, look, I get it. I'm A, I'm an it nerd too.
18:07
I've spent an entire week training on a, we won't name the vendor, previous vendor X that I, you know, used to, uh, used to use when I was a customer II, I get it. And, uh, you know, sometimes you feel like, oh, yeah, it's fun. I'm the guy with in the know, but how much, how much more valuable is my time to go and apply it to other parts of the
18:25
business to actually deliver on what, what our differentiator is, right? So, yeah, absolutely. If any RO remembers vmfsvmfs alignment and the guest file system alignment you have to do inside, you know, some of that it's enabled by things we'll talk about later there. There's also the, the takeaways out of this is this, for instance,
18:42
um this actually goes back a while. There was actually a customer where they had, you know, half their storage team, um act, they actually had half their fleet on pure, half their fleet on someone else just being very real of that. Of those folks, one person managed uh the peer devices and actually he spent maybe half his time doing that and the other half doing automation thanks to API first stuff and then
19:02
the other, the rest of his team, it was more than four or five folks actually managed the other half of the fleet, right kind of thing. And this is even a little bit of a career coaching mode. If, if this is what we do day in and day out, this can sometimes be a little bit challenging and even a little bit scary, there's also the fun, you know, graphic and this is still originally the,
19:18
the install for pure was a, was a tent card. We don't have some longer manuals. Right. But fundamentally, you get the system up and running, you can actually fit it on that. I've actually, I should have been good. I've got some of these in my desk right down here. I'm not gonna try and get them out on the fly.
19:31
You know, shame on me. I think it's so much fun that we still sell those in our company stores. They're still there. Yeah. Next consumer flash and ML consumer flash and data reduction and we wanted to kind of bundle these together as ideas just because if you look at even 10,
19:51
12 years ago, MLC was the consumer flash of the day. Oh, it's probably QLC today now we're gonna go there. But the idea of how do you find the cost curve and ride the cost curves that make flash viable for customers, whether you're using consumer flash and knowing that the production of consumer flash was gonna insanely outpace. What at the time was Enterprise Flash,
20:10
right? You know, because we look at uh the iphone and the ipod and all that stuff that generation, right? As well as then how do you actually, you know, if you say it, it's on flash, how do you actually embrace what flash is in doing some neat things with data reduction? Actually, I'll just pause here over to you
20:27
again. JD. Yeah. And, and you know what you kind of alluded to those earlier days when we were looking at the cost curve of hard drives versus flash. You know, I think a lot of us, a lot of people thought we were crazy when we said, no, we can actually build a system using flash that it can be cost competitive um for the right workloads against disc. But what we've seen as that story has evolved
20:49
and evolved and evolved is that the the evolution of the media types that are being made available to us continued to prove that investment out and I'll give you another example. Um Do you know Andrew the largest enterprise hard drive that you can get today and I know somebody's probably gonna find one in the chat that's a little bit bigger. But I think the largest enterprise hard drive
21:10
and a quick little Google Google search. I could do got a guess seven bazillion petabytes. Not quite that would be handy 2020 terabytes. What's our, what's the biggest flash module that we now have in our products working on one terabytes? So yeah, so I mean, you know, that early bet in how we able are able to use that, you know, variety of media types.
21:37
It says consumer grade here, I think consumer was, was really a brand name of how that flash was used at the time. And we really proved that that wasn't necessary. They didn't have to be the truth. We could actually take that media grade uh TLC which was called consumer. And we could actually, if we treated it the right way,
21:54
if we understood very deeply from an engineering level how it worked, we could actually give it enterprise characteristics while bringing down the cost. And so that's a big part of of kind of the early genesis of what we were able to do to kind of make this whole story work out, right? And that then pulling that forward to today, you know what you and I were talking about last
22:13
February. This is actually a part from that exact same one applying those same principles to QLC. A lot of the same engineering work around that stuff treating flash like flash, you pair that up as well. And what it's paired up with is also data reduction. So this is where um it was fascinating just digging into the inters of here when I joined
22:29
about how in line ded duplication and compression, post process ded duplication and compression. It's variable under the covers a lot of stuff I used to think about different architectures that leads to, you know, better data reduction efficiency at scale. In some cases, it actually changes even the way that we can do cashing or not needing front end,
22:48
right. Cash inside, you know, the controller we make sure to protect against if you, you know, if you lose power, you shouldn't lose data. You know, that's even where it always gets illustrated this way too, you know, pattern Moval de compression, deduction, copy production. All of this stuff was actually built in from
23:02
the very beginning and that's just kept paying off in continuing to bend the price curve of flash to dis even closer and closer. And I, and I'm pretty sure daddy, we haven't run out of runway yet for what we can do there. Absolutely not. You know, I, I think there's a, there's still a lot more that uh that, that I look forward to.
23:20
How about that, we, we'll leave it there. But one of the things I want to highlight though, I know you're about ready to move on, but we think about data reduction and usually the first thing to come to mind is that's a cost saving, right? If I can store more data in less space, I have to buy less and then I save money. And that's absolutely true.
23:36
But it's also really important to highlight it's also a big performance benefit because think about it, if I, if I'm literally storing less data sets, if there, if, if I've reduced that down, then all of the ancillary operations that I have to do on that data from reading and writing to all of the other metadata management and everything. I, I'm literally doing that on less smaller data sets.
23:57
I can be more efficient faster and I get a performance uh benefit as well. And so, you know, it's, uh it really is kind of one of those win-win scenarios, right. We've actually seen some really neat things both out of that as well as the, the failure rates on our direct flash modules being dramatically lower. I'm not sure if I'm what I'm the numbers,
24:14
I'm supposed to say in public just be very real. But when you look at some of what peer does with keeping flat and fair maintenance and not spiking the maintenance after year three, it's not just a financial play, it's actually driven by some of these capabilities leading to lower failure rates. So it works out for pure and for customers. You know, that's like a perfect world next third kind of item to put it together
24:34
statelessness and the controller architecture. So this is where and we even referred to this heavily about five months ago about stateless controllers that the controller, there's nothing there that's unique, inherently unique to the controller. Everything is virtualized as far as a controller or architecture standpoint like the WWWWPNS and other pieces,
24:53
the controllers, the NV Ram is outside of the controller that eases, that's a big one. Keep going. No, you think about it. You know, typically with a lot of systems that NB ram is, it's stored on a controller and then it's maybe copied to another controller. And it's very difficult in, in a lot of architectures to be able to remove one of those
25:14
and, and, and, and lose some of the resilience of the state. So I think very early, you know, I don't know, I just kind of geeing out here a little bit talking about random stuff. But, you know, one of the really early architectural things that we figured out was how to essentially dual port um memory so that I could have two different controllers, two different devices,
25:33
talk to the same memory. And that, you know, that one engineering breakthrough has had profound impact on the ability to, you know, to do some of the things that we're talking to have stateless controllers, right to, to, to deliver on this promise that paired up with very intentional. Now we're talking about flash. Let's be clear here.
25:51
Flash play is a different scout architecture, an intentional dual controller architecture where we're reserving performance capacity. Like say what? Well, if you actually want to make the promise of having non disruptive hardware update upgrades, even between families of controllers, you need to have some intentional backend engineering to make that possible. There's some other cool stuff here about not
26:09
needing multi pa software and the balancing across ports and you know, and tuning, tuning and balancing. So there's not by default impact if there's even a software fail over for, you know, a software upgrade or otherwise. So there's, this is one that often we'll park on sometimes when we're doing deep dives JD of like this is the front end, active, active, all the ports are active,
26:28
there's even NPIV to mask the WWPNS. And so you don't have to have a wait for path, pathing fail over from an NPIO stack. But the backend active standby, there's some real intentionality there to deliver that promise of non disruptive generational upgrades. It's non disruptive also, in terms of the performance profile,
26:47
right? If you're looking at what is the capability of the system, if I, you know, suddenly remove half of that capability, am I really fully online? I, I would argue you're not. And so, you know, being able to, even during these, you know, migration or failure states being able to maintain, not only the availability of the
27:05
service, but the availability of the service within the slas that you're expecting. I, I think that's a key part there. I think I'm gonna make sure that I'm watching time here as always. So these couple I'm gonna kind of, I think I'm gonna blow through a little bit and then I'll, I'll toss it back to you one more, one more time, streamline code paths.
27:23
Actually, I'll try because there's just so much good stuff here. This is the idea that um that any, any code path that's not part of the normal system function could be broken. The best example I know of here is how often do you go and power down in a storage array? Not very often a lot of systems, you've got this manual process and do these things or
27:40
maybe you have to have someone come in and certify it and they move it kind of thing with pure. It is literally the unplanned power off process is the same as the power off process, right? Planned power off process. It's yank the power cable, we do stuff to perfect and make sure we don't lose any, right IO et cetera.
27:54
But the idea and even we actually exercise our raid code every time that we do a read, we actually try and read the data as well as rebuild it from raid at the same time. And it's a race to see who's faster. That's because of some of the characteristics of flash that that and even limited air pad sign comes from a back end hardware standpoint. I'm not even gonna pretend I remember because I don't actually who it is on the right hand side,
28:14
but usually look at pure systems. They are one third to half to quarter or less of the racks, units and systems and parts and pieces. Um There's even an analogy here about, you know, you take two and four engine airliners actually, when you work out the number of components in the M TB F for large systems with lots of moving parts. It actually doesn't necessarily work out.
28:32
Oh, there we go. So a Polish um last, then I, we aren't going to talk much about immutable snapshots. I do want to note that that has been part of the core architecture for a long time. We'll talk, we've hit on that on ransomware ones. And then the last one here is security encryption, fun detail on pure encryption has
28:51
actually always been on from day one. And it was actually partly because we leverage encryption to randomize the data for ware leveling on S SDS. That's part of why you can't cut it off actually. Yeah. What kind of workloads are, are, is flash really, really great at random workloads.
29:07
Well, let's make everything random and ware leveling that goes into the, you know, the pieces that you saw before with handling MLC and QLC. So this even goes back to the origin always on encryption as well as and not just a oh, there's performance overhead. You need to check a box for, you know, agencies or certifications. No, no,
29:24
it's actually a core part of the architecture in some kind of clever, clever ways. So hopefully this gives everyone a little bit of a sense of there's deeper stuff under the covers here that go way, way back. And when you see a slide like this, it's the same one you saw just a little while ago and it says you know,
29:44
97% still in production today, you know, and you can update, you can do non disruptive upgrades. It's like, ok, that makes a little more sense. It's just not bold claims that I'm not, I'm not sure if I believe so. Let me highlight that. I, I just wanna, you know,
29:59
we, we, we talked around it and I think everybody gets it, but I just wanna, I wanna make it crystal clear what we're saying here because it's so fundamentally important. I and I have these toys that I wanna play with. You could literally have started with the array that has this face plate and you could have through these evergreen uh um updates, you could literally have
30:21
upgraded that system, both controllers and flash, even the face plate. Of course, all the way to the latest generation of hardware, your array may look like this. Now you've never taken an outage, you've never migrated data. You've never had to worry about planning for your applications working on the weekends because you had to do some kind of migration.
30:41
You never had to worry about any of that because non disruptively we converted from the first array all the way to the current array media controllers, literally absolutely everything. And that is the value that I agree. That's the what you get from stateless controllers and some of these other architectural decisions.
30:59
There's a great comment from Keith in the chat. I'm never saying last names just to be respectful in case, you know, uh, he'll watch in the future, but, you know, noting that he's actually had to shut down his data center in the last several times in the last few years. And the pure system, you know, this part of the way the pure A is definitely the easiest to shut down and start
31:15
up. Cool. You know, I love hearing that. So love it. Hopefully, in some cases you don't have to, you've got to do a full data center shut down. You know, of course, you gotta sit down everything. OK. So let let, let's let's switch gears here a little bit. So there's this underlying uh foundation
31:30
architecturally, we talked about MVP and kind of design center. We're 12, 13 years away from the very original flash, a design center potentially at this point. So it's like we're, we're stretching from a time standpoint, but from a hardware and software standpoint, we're still actually, in some cases realizing the benefits and playing out the benefits,
31:52
the potential if you will. That's why it said actually the potential flash rays foundation playing out that potential from a hardware standpoint. I think there were two pieces that we wanted to hit on. It's bezel time JD, right. You, you, you, I had the flash, did I jump ahead a little bit? Yeah.
32:09
The so flash ac Oh Thank you. And you can actually see how it lights up with a little like a little, little orange lo around that kind of thing. Flasher Ac G personality coming totally through the vesel. This is what we talked about last year. Uh There's been some updates since then. Am I taking this through? Yeah, absolutely.
32:28
And so this was, it goes back to the, the we called it consumer grade media but that it didn't stop there. The, the industry continued to innovate, continuing to give us new technologies. And one of those new technologies was something called QLC and QLC is about storing even more data in the same physical footprint of a flash chip. And so, um you know, we were very, very well
32:51
positioned because we did a lot of work early on in making traditionally less expensive grades of flash work at an enterprise uh capability. We were very well positioned to take that new media and put it into a system that was able to provide a really a shift towards density, still very high performance, still flash, but really focused on giving you petabyte scale density into uh a much
33:19
smaller footprint. And and it's really allowed us to kind of grow, you know, a lot of people, especially in the early days thought about flash ray as for those, sometimes we talk about them in terms of tiers, tier zero workload, those ultra high performance workloads. But with some of the evolution that we have, we're able to start addressing workloads across the data center.
33:37
So now, we can truly use pure for anything, any workloads, any needs across the data center. And that's really some of the value that we see here. So being able to, I'm just kind of going back to my notes here. Being able to uh expand granularly was really important when you've got really big data sets, really big amounts of storage, being able to do a little more granular upgrades is something
33:57
that we introduced uh in rev three here, um improving the data reduction capabilities of the system learning more about how QLC works and optimizing our code again, optimizing on our side, not giving you all those knobs to turn. We're actually using the analytics and the intelligence that we get from these systems in the field through pier one to go and make those decisions for our customers completely
34:19
transparent to you, right? And so um ultimately allowing, allowing you to do more with uh with pure and with flash ray in your data center, you may have noticed the uh internally sometimes as we call it, you know, the generations of controllers and hardware. So we added the, we had the C 40 we updated the C 40 also added AC 60 you know,
34:36
made it a little bit bigger or maybe I had that backwards. But there was an additional controller, a controller option. We've then also gone and done some stuff at the upper end. So if we look at, you know, and this is where we'll talk about kind of the the flash ray family if you will.
34:49
So flash ray X has been around for a while or the predecessors of that we've been talking about flash AC we also added at the upper end flash array XL. Pushing further from a performance standpoint inside a single performance envelope. We had to do some interesting things with um handling thermal envelopes and thermal offload. I mean call it more fans, but it's fancier than that.
35:08
It takes hard work with processors that can do more that pushes further from a capacity standpoint. Also from a performance standpoint, especially around high throughput workloads. You know, you can actually see the kind of the relative numbers these are comparatively accurate. Now any any given workload you put on, of course, your throughput varies based on your
35:25
block size and rewrite mix and all that kind of stuff. But these are relative numbers that show show that. So flash ray XL allow us to push further in single larger workloads or large arrays. We still see customers with fleets of P arrays and we're doing pure fusion and some other neat things you see that to handle with like fleet
35:44
wide management. But it actually those same principles about flash array X and flash array C. What you'll not notice on this slide is oh it's new software or it's new data reduction code or its new replications like all that stuff is the same and even some of the neat stuff that we did with uh implementing our same NVRM architecture.
36:03
If we didn't tell you that we'd done some hard engineering work about distributed NV RAM architecture that actually matches some of what we did in flash play previously. You wouldn't even know it gives you the exact same resiliency. So there's a level here of it's all the stuff that you love about flash array and maybe about flash ray C depending on what product you started with the for your family,
36:19
but actually taking it further and bigger as well. Yeah, and you touched really briefly on fusion and I wanna come back around to that because let's be really clear flash array XL absolutely has the same code base, the same purity operating system that flash ray X that flash ray C does. So if you're one of our customers today and you're already using flash ray,
36:40
you know exactly what this is gonna look like from A U I, from an API perspective, from a feature functionality perspective, you know, you absolutely, which you're gonna get there. But one of the ways that this plays into our kind of overall strategy is pure fusion thinking about how we help customers. You know,
37:00
we we we did a lot of work to modernize infrastructure to make storage better, but we want to go further than that. And as we kind of look to the future of some of the products that we're developing like pure fusion. We're really thinking about how do I modernize operations? How do I, how do I give you that true, cloudlike experience operationally,
37:18
not just and, and it's important, the cost models and consumption based uh and all that stuff is really important, but we actually want to think about operationally. How do you make that incredibly simple and having, having a lot of different classes of storage plays really well into a very large system like Excel. And so there's lots of fun things that we can do there software,
37:43
you may be guessing if you're listening to this. So there's the Mark Andreesen quote about software eating the world. Uh that was too long for me to actually fit in here. So, you know, software world eating but flipping around talking about flash ray software because in it fundamentally Flash Ray is a software platform,
37:58
right? We leverage the same software on hardware, on different hardware platforms, even in a totally software environment in the cloud with cloud blocks store, right? That kind of thing. So if we look recently at flash array software, um you know, actually pulling this same slide up again, you'd be like really that one again.
38:14
Well, it's like another piece of that it says here, you know, 20 plus software upgrades. It's always fun to pull this up, especially when I'm signed as part of our jobs. Um You and IJD will, will do road map presentations for customers, which is hopefully more about kind of vision and strategy and where peer is going with a
38:29
with a future look kind of thing. And so usually before I talk about anywhere where Peer is going in the future, it's for any given person that I talk to, they may not realize how much has already happened over the last eight or nine or 10 years with this just steady cadence of flash ray updates. I, I think one that I find the most interesting here because there's a bunch of stuff here.
38:49
This is a little bit of a, um, little bit of an eye chart for you is the deep compression one. we actually added in some new deep compression, post, post, uh, post process compression libraries, uh, back in the 51 days. And even before that, where they were actually, you know, enhanced, they were actually sometimes customers that they upgrade to a purity release
39:08
and then they actually had less used space on their array, which is a little bit freaky. Like you're like, where's my data going? It actually was because the crawler was going back underneath the covers and recompression data more efficiently, which is like, that's really cool. Once I know what's going on,
39:23
I think you want to, you want to highlight here JD. Well, yeah, like we, you know, we, we talked about how customers who started at, at the far left of this chart through that, you know, stateless controller architecture could have migrated all the way to the end and, and software plays into that too because, you know, think about, think about a traditional purchasing model where you buy an
39:46
array and it has a certain amount of capability or certain set of features that it has. And, you know, maybe you'll get a couple of software upgrades, but eventually you're gonna run into a wall when it comes to hardware, you don't have the capability to do the fastest networking architectures or, or new protocols that are available or enough epu to be able to actually process the IO
40:07
required for something. Um A as, as impactful as active cluster, right? And so I want to highlight that the customers who have gone through this journey with us and non disruptive, upgraded their hardware from, from version to version also have been able to take advantage of all of these software advances that we make. And how cool is that you could have purchased
40:26
an array in 2012 and that same array maybe not with the same components because you've got gone through this upgrade. But that same asset, that same investment now is taking advantage of hardware and software capabilities for 2021 and 2022. And I think that's phenomenal. A couple of recent ones to highlight on flash A uh there's a wealth of replication capabilities
40:49
uh included. Uh some of these didn't exist, you know, a couple of years ago, synchronous replication. Um I've worked with some other synchronous replication technologies. I'm gonna leave the names out but I never would have described them even when I was like whiteboard and architect them as simple and included.
41:03
Mm Just not what they were, right? Extra cost, extra complexity, et cetera. We took that further within about the last year and a half what we call active Dr which is instead of synchronous replication, which is usually about 100 miles give or take active Dr says, you know, can have RPO in the low minutes even.
41:18
I was on a call uh earlier this week with a customer who was testing active Dr and had it at four seconds and it's not a promise. It's what they were seeing in their testing. It was really cool to see the RPO in four seconds. We then went and now dialing in specifically on the most recent flash A A 6.3 release. We had a tech talk on this just a couple of weeks ago.
41:36
So if you want a full session on that, feel free, but or make sure to go back and look at that on the pure storage dot com slash webinars. But there are a couple of things to highlight. We took that active DR capability on the block side and we extended it over to the file side of flash ray, flash ray is not just block only, it also does file now as well. File is not easy and we're investing in it. And this is the one of the ways that we're
41:57
investing in it. So simple setup, similar to what you'd expect from pure, follow some of the pod construct and even some of the easy fail over capabilities and testing capabilities, etcetera, that kind of thing. Next major item in 6.3 upgrades. II, I think you actually almost had a contrarian take here JD on this whole idea of uh of self-service upgrades.
42:20
Yeah. You know, I, I just, I, I wanna point out we've, as we talked about, we've got a long history here of, of flash rates, gone through uh six major firmer revisions, multiple hardware revisions, uh lots and lots of upgrades. Uh W why, why um have I up until now had to go to support if I
42:40
want to upgrade firmware? I mean, we've done so many hundreds and thousands of these at this point, but I'm sure we've got it figured out. Right. Yes. So the even piece there and actually pardon? One second here is um ok, there we go. So it, it, it is. And the interesting thing there is that we have
42:58
um I can look at it as like, hey, should we have had this sooner? The focus on pure for a long time has been around what we call white glove upgrades. Support is involved. They're running scripts, they're checking health. I mean storage matters. We don't take upgrades casually. The the non disruptive part has to be capable
43:12
from a software standpoint, but it also takes intelligent people doing smart things and checking the things that connect to the pure device and running strips. So what we've done is as we've continued to grow frankly from a, you know, size of customers and number of flash rays out there and even what customers want is they have larger fleets you saw on Pier one a while back there was actually self service.
43:31
Um There were upgrade scheduling capabilities. This is now taking some of that capability and putting it into our customers' hands if they want it, not saying you have to do what we use to help you with. We'll still take that exact same approach. But at the same time, we're pushing further from an automation standpoint based on what we've learned by doing
43:47
so many of these upgrades over the years. Yeah. And what I really want you to highlight and you did is that, you know, this is not, we think about upgrades, I download a bit of firmware and I slap it on a device and I reboot it and it's good to go right. So much more than that, all of those uh those those health checks, those at that white glove service that you talked about making sure that
44:04
we have an incredibly high degree of confidence that when you go through this upgrade process, it's going to be effectively seamless. And so it's not just applying a patch and moving on with your day. There's actually a lot that goes into it. Right. I'm gonna ask indulgence for folks. We are right at the 45 minute mark. We're gonna go just a couple minutes over.
44:20
We'll still do the drawing. We'll still get you out of here within the hour and we'll do some Q and A because the last one that we wanted to put in here is actually when you see this a little bit of a bold claim, but it's actually more the vision of where we're trying to help our customers go to. And that sometimes and in a space that is frankly really challenging city.
44:43
Yeah. And so, you know, I I we, we, we started with this capability called safe mode and we've had it for a while now. But what I want to highlight is safe mode is not just one thing, it's not just I can lock Snape snapshots from being deleted. Certainly, that's part of it, but it's really a progression.
45:01
And what we think the way I like to think about this is in the, in the early days of some of our safe load capabilities, there was very much an opt in for the customer, right? You had to go do something that wasn't complicated. We tried to make it incredibly simple, but you still had to go take some actions Right. And, and I,
45:21
as many customers have taken advantage of that. It's the saddest thing in the world when you, you have that unfortunate scenario where the customer has the capabilities, they have the features sitting there on the ray, but they just never opted into it. So something bad happens and those, those snapshots aren't there, they, they don't have what they need. And so as we progress,
45:38
maybe I'll kind of let you talk about what we're moving towards. But, you know, we wanted to kind of improve that experience a little bit, right. It's really a question of how can we help. So even in some cases, we talked with customers that have done a lot of security investment and events and depth and Attackers are smart and capable.
45:52
So how can we help by starting to shift toward a opt out versus an opt in, in that add in some level of granularity? That's where it says flexible without compromise and use case granularity. There's actually actually some granularity at a protection group level versus full array level. If you're gonna do so, we're trying to help our customers go further, not to today,
46:10
you must do this. But when we see challenges, even if it was someone as a conscious customer choice, if I don't want to use this feature, because, hey, that, that is fundamentally your choice as a customer or it was my choice as a customer. But how can we help by having the protections, more default, more automatic so that they're there when you need them versus when you
46:29
realize you need them. And that's usually too late. That's just how these things work kind of thing. Unfortunately. So with that, I think JD, we are, we are right here. Um, at the end, I mean, the, uh, the answer is 42. Now, what's the question we wanted through? Just a ton of stuff as always.
46:46
Um It feels like this flies by. Thank you so, so much for joining me. Thank you so, so much for having me back on. I always have fun with you, Andrew. Likewise. And for those of you, if you enjoy this, by the way, uh JD and I actually hopefully did,
47:00
you know, I, I think just about everybody stayed with us. I didn't see people drop it off because we could watch that, you know. Um JD and I actually joined Rob Ludeman on a kind of a, oh, I don't know if it's a subset of the report. We call it unplugged. It's where JD and I go back and forth about
47:13
stuff. We're seeing, we actually go through the 15 decisions in more depth. I think we're on number four, we're taking a lot more time on each one. So, um if you enjoyed this, uh go check out the Pure report, both in general but also the unplugged, we're up to, you know, four episodes. We're gonna be recording another one.
47:27
I think we're supposed to do it next week. If we stay on target, you know, kind of thing. Finally, please make sure to join us next month. I'll be back with Jack Hogan. We'll be talking about what we're seeing from cloud trends, public cloud, private cloud hybrid cloud, how you can make that fit your business.
47:43
And if you're waiting for the drawing, here you go. So I want to thank Brian S from Houston for staying around with us. I hope you did. Uh Emily will be reaching out for, to uh provide this to you and ember mug and you can actually turn it on out with. Um No, I don't think he want this because I've, I've been drinking out of it as you know,
48:01
people notice, I mean, we could sterilize it but no, no, no, it'll be actually one that comes since I had this sitting here, it'll actually come like in the box, you know, look at that. This is, you know, so much more attractive than seeing me. So, you know, less attractive, more attractive. Yeah.
48:15
Oh, well, you know, with that, if you need to go catch a bio break or eat some lunch before your next meeting, please feel free. But I think JD, this is where we just hang around for, you know, five or 10 minutes and we kind of turn the music up a little bit of sweet jams, hang out for Q and a let me make sure I don't get that too loud here.
48:35
Pull it back to the beginning. Here we go. This is now where for if you're staying around, uh please feel free. We are out of, I don't think we're ever super formal but you know, we're out of like we're going through stuff and now it's just actually looking at some of the questions.
48:51
Was there anything JD before? Uh before we look at the questions and Sean and Jeff and Zain? Thank you. Anything that uh you couldn't squeeze into the time that you wanted to make sure to highlight. So, back to you first, I think Jeff. Absolutely. He had, he had some comments on our internal chat about object granularity when it comes to safe mode.
49:08
Jeff did, did you wanna jump on and address that or did you want? I don't think I have the level of uh technical uh technical uh depth to get into it. Yes, it, it does address the concern around uh data usage problems if you want to go out a little bit deeper, Ken. So in general, uh safe mode helps protect. So there's snapshots that gives you immutable copies of your data that are pointer based that
49:33
don't take space safe mode helps protect against malicious data deletion via a two step deletion process you destroy and you eradicate that goes back to the origin of pure. We saw that maybe when we flew by the snapshot slide. Um What actually now we're adding in safe mode in 6.3 is the ability to make the safe mode, data protection, data deletion protections more granular than being on the entire array.
49:56
You can actually do by what we call protection groups. You put volumes in protection groups. That's where you also create the snapshot schedules about how often do you have to create snapshots and how long do they retain for on a granular on a, you know, a group by group basis kind of thing? A safe mode can now apply at that level or at a full array level.
50:13
Candidly, I almost want folks to more use the full array level unless they have a specific need, it sometimes around backup interactions, et cetera. I'd rather you go with more default protection than granular protection, but we recognize the need for that, especially as we go to auto on safe mode pity. Yeah, just another example of how, uh you know, Evergreen makes these things better with time.
50:35
And we started with a really awesome core capability, you know, actually several years ago with uh just the concept of a dual stage deletion and eradication. And we've just been able to, you know, and build on that investment all the way up until now and we'll continue, right. We'll continue to find new ways to make this an
50:54
even better uh capability for our customers. Ransomware is not going away. We're going to continue to have to give our customers options to, to to defend against it. There's a great comment as well from Samir along those lines. You know, we're looking at implementing safe mode and expand on the issues.
51:09
The other thing I always make sure to toss in there when I go through often anatomy and attack with folks is I think someone is a complexity landscape, some of the other pieces or options that are in a ransomware protection, especially from someone compromising admin level credentials. That's hard to think about how to protect against.
51:26
There's a huge amount of complexity and cost there. So both you and I did, we spent a lot of time digging into what I think it was like interaction scenarios and there's some complexity challenges there. But compared to the alternatives in this space to protect against those kind of scenarios, dramatically simpler kind of thing. So maybe that's talking about both sides of my
51:45
mouth. But, you know, I was thinking about what other options do I have. Um There was a, a question here uh from David Silver Horn. Uh Maybe I'll toss this over JD about um in the past that pure had, you know, windows server kind of windows built into flash array.
52:03
And uh we're not doing that approach. Uh You want to walk through a little bit of that history or I can either way, you know, I, I can, I can start it off, I think you actually have a little bit more history here. So maybe you can kind of embellish after I start. But yeah, absolutely. So, again, another illustration of how we've
52:17
kind of matured the pot product capabilities. There was a period of time where we actually had the ability to run a, a file gateway, right on the, the flash rate. Is that what we're talking about here, Andrew, I can't actually see the question that we could actually run a windows based file gateway. Now, there's uh that, that was great because it,
52:36
it, it unlocked a use case time that we couldn't support. So it gave us a really early ability to kind of do some work in that space and offer some file services. Uh But as we've matured, you know, we, we, we know that there's challenges with a gateway approach, right? You don't get a lot of the same efficiencies,
52:53
you essentially um hide or mask a lot of the benefit of flash array from a a data protect a data reduction perspective, some of the data service capabilities, a lot of that kind of gets kind of hidden behind that gateway. And so as we've moved forward, uh we've actually worked to integrate file services natively into flash ray. Now, uh I'll admit that's for it's for certain
53:15
use cases. So reach out to your local team and make sure that that's a, a tool that's going to deliver on the use cases that you need it to. But when it does, that is a way that we can kind of get the best of both worlds. Again, we get those file services natively on flash ray, but we get it without having some of the limit of a gateway approach.
53:34
And, and again, it was, it was awesome when we didn't have native file services that got us to this point. Um So that, that's cool that we had that, but we're really moving more towards having those native capabilities. Uh It allows us to, you know, as Andrew highlighted active Dr now for files, we can actually natively replicate uh files with active Dr.
53:55
And so I think that's going to be the approach moving forward, I think all packaged together two last ones here. And uh so there's a question about, you know, safe mode with the VM Ware Srm and even uh cos rubric making snapshots and integrating with pure. So the, the uh per safe per protection group granular safe mode capabilities actually goes
54:16
to the ability for backup software that might create maybe ad hoc snapshots, you know, called from an external API to actually be able to remove those as well and then save the snapshots, the temporary snapshots right? Short lived. Yeah, they're not meant for long either. And when I say long term recovery, in the case of ransomware,
54:33
you don't want to go back weeks or months, you want to be T minus one second before the point of encryption, that's usually within a couple of hours kind of thing frankly, right? Because we can't roll organizations back a month. We all go out of business with the systems of record related to that great question about safe mode and site recovery manager.
54:48
Um Today, those do work together full disclosure. There is more work coming in the next couple in the next quarter or two. John, if you have what need specific timelines there, please feel to reach out the fundamental question. This is one interesting one to think about. Is that anything that is deleting data or removing data from the array to destroy plus
55:06
eradicate process? We have to think about that through the lens of it's being done for good business reasons or it might be an attacker who might have compromised the site recovery manager server or the backup server. And I'm not calling out our backup partners at VM Ware, right? They just got credentials to that and they're
55:20
trying to do the same operation. So it's that thinking about how, how do you approach for systems that may in site recovery manager's case, want to clean up from a disaster recovery test? It's one of the core values of SRM. And how do you map that into a safe mode approach? So there's already some some stuff in a really long great KB I want to say by Alex Carver that
55:38
actually outlines how you can handle that today and then there's some continuing improvements coming in the future. But there's a fundamental question of if things are getting cleaned up or deleted, we have to think about what if an attacker got access to the same credentials that, that external system has. I think with that, the music in the chat say they were hungry.
55:59
So, I think, uh, I think it's lunchtime, at least for some films. But I, it is, um, I, I think, uh, you and I are alike would periodically get later lunches than we'd want to. You're on west coast time. So it's not too bad for you. I'll agree with you. I'm hungry.
56:15
I'm looking here for any last questions. Any last thoughts, think the, um, and actually I'll put, uh, there, there's a great comment in chat from and actually I'm just gonna make sure to call us out for Doug. I've actually had it. Please reach out to pe we have a plan around any customers that are running windows DFS the way that you are.
56:36
Uh, we are not discontinuing support. We are still supporting it for a good long time. There are capabilities there. We are not leaving customers behind. It's a little bit of a different flavor, frankly of evergreen just because it's a different thing. But that foundational piece of pure, not leaving customers behind in a jam. So I've actually had some,
56:53
uh, in depth discussions with customers around this who've had this exact question because we haven't been, um we haven't been private about we are investing in a unified capabilities where we're going longer term, but we recognize we have customers who have used some of the stuff that we've already had out there. So please make sure to reach out to your se uh you may end up having a principal technologist myself, JD,
57:11
Sean Kyle, all the other folks on this team that may be pulled in to help with that discussion for you for a path forward. Yeah, great, great, great pitch. Thanks for highlighting that one, Jeff and with that the final one for those who are still here, I always appreciate. It. Seems like we always have a ton of folks that
57:27
hang out. Uh You've got the QR code to register. You'll see the link in your email at some point because hey, you've agreed that we can email you about the coffee breaks. Thank you for that. Please make sure to join us next month. As always JD. It's a pleasure.
57:40
Thank you till next time. Thanks everybody.
