00:00
Uh, right, hello, everyone. Um, we're going to talk to you about how a pure customer is using Pure Fusion to wrangle the problem, the problems of growth. Uh, I'm Larry Tuche. I'm a director of product management at Pure. Uh, this is the customer. This is Mike. You want to say something about yourself?
00:19
Yeah, so I'm Mike Dehaan. I'm the infrastructure services manager and enterprise architect. Yes, my job title is actually this long at the Mississippi Department of Revenue. Um, have any of y'all heard anything about this new thing called Fusion lately? Um, I feel like there were some words about that in the keynote.
00:35
So we've actually, I've been working with Larry, uh, through the products on this for a while, and we actually deployed some workloads with it 3 weeks ago in our production environments, and we're going to talk about how it's been helping us with our problems of scaling. So I think one of the main ideas we're looking at is how big since we've grown since what I call BP before Pure.
00:56
Uh, we've grown 1,800% in our storage environment and fusion is really helping us wrangle that. All right, cool. I'm gonna start off, uh, you, you're actually gonna, you'll get to spend most of your time listening to Mike in this session, not me, but I will set the stage and add a little bit more color around what these features are doing for Mike to kind of drill into what you heard in
01:18
the, uh, the keynote just now. And so the problems we're trying to solve, uh, with Fusion and everything you're going to hear. mic using is uh in the product today. So this stuff has actually shipped in the flash array a couple months back in February, and um some of the technology you'll hear Mike talking about will come for file services on
01:40
Flash ray and Flash Blade later, but a lot of Fusion features already work across FA and FB as we go through here, I'll try to make it clear what works where and what's coming. Um, but the problems we're trying to solve here is that um if you're all pure storage users today, you know that each array is an independent management endpoint. If you want to do something to a volume or a file system before you approach your browser
02:04
window to get to the UI, before you use an API to go manipulate that thing, you have to know what array that thing is on so you can connect to the right endpoint. Um, so the arrays themselves are siloed. um, that's also that, you know, that's a narrow. Uh, a definition of um how manageability works, uh, today,
02:22
but it's also a broad, uh, thing like what Charlie was talking about about how the entire stack being siloed. Um, you also have a lot of uh resources to manage, and what I mean here is when I say the word resource, I mean the volume, the file system, the production group that configures snapshots, the pod that you would use for managing synchronous replication,
02:45
uh, the access policy that you would attach to a file system, those are what I mean by resources. Um, all managing all of those things in Pure today is very simple, but the more your environment scales and the bigger it gets, and all those little options stack up and you have to do them all just right.
03:01
If you don't, that'll lead to mistakes and risks, right? You forgot to put a volume in a P group, you forgot to set the retention that you wanted for your policies on the P group, that sort of thing. What um, what we've done to introduce this problem and to wrap like a new layer of simplicity around managing all these things is introduce pure fusion.
03:22
Um, I'll run through this very briefly just so that you have a high level understanding of what it is. Fusion is simply a purity feature. When you get on the right version of Purity for Flash ray or Flash blade, Fusion is in that version, and you can turn it on and use it just like the replication feature. It's a couple of simple clicks to get it set up and turned on to use it,
03:42
you don't have to deploy a virtual machine, you know, it's not an external control plane. It's a. So we're there somewhere else. You don't have to connect your arrays to the cloud. If you do, you'll get uh an enhancement in, in terms of recommending which array to place a workload on. You saw that in a demo on the main stage.
04:00
We do use the pure one cloud to do that, but we just use the phone home connection to make that work. Um, OK. And so the idea here is that what we're working towards, and we've delivered this for block workloads on Flash array, and that's what Mike is using. Is the ability for you to define the outcomes you want.
04:18
That's using a thing we call a preset, hand the work of deploying the workload correctly over to Fusion, let it do the work for you to provision all the stuff needed, and what you get out in the end is a consistent configuration that complies with all the standards and policies you want to have in your environment. And that is it for me. So until the end, I have,
04:41
um, I think I'll give the demo and yeah, OK, yep, all right, we'll let you hear from Mike. Oh yeah, we need to trade places so we can, um, so our story with Pure and my story with Pure actually personally, um, start at about the same time to the day. So paint the scene for you guys.
05:00
Uh, I'm fresh out of college, right? While I was in college, I was the IT guy at a local copier company, the only IT guy, right? So I know for anybody that's ever followed a printer before, that sounds like to set up for a bad joke, but I swear that was my lived reality.
05:15
Um, so we go in. So at this point, we weren't running any of our front facing applications. We're running a small little 4 node HCI cluster, and we wanted to take on ownership of our front facing applications. We had another agency that was running that for us, um, and we thought we could do better.
05:32
Uh, so we got our first flash array in 2016. It was such an enjoyable time for us. We're standing it up and I was drowning, um, right, because at this point I had never worked on any enterprise systems, but then I looked at the CVD and started reading it, right? Cisco Validated Design was my gospel at this time and everything started to click and kind of makes sense,
05:53
right? And so we got our application up, our first app in. It was awesome, right? Our business units loved the performance that we're getting and all of that. Um, so as we're kind of rolling through that, we start scaling, right? because our business units loved it,
06:05
they wanted more systems to get put on there. Uh, so we started moving into, you know, more arrays. We did a non-disruptive update. Uh, we've, you know, scaled our arrays up, and then we actually got this one right here. So this is just one little bullet point right there, um,
06:21
that hangs off the backside of the longest Monday of my life. So has anybody in here by show of hands ever had to use their DR strategy? Just curious. OK, so we had to use our DR strategy because we had data corruption on a Monday morning. And we had a purpose-built backup appliance at the time that I am not allowed to say who it
06:43
was for legal reasons, and it didn't work, right? So we get to the end of it, we finally recover all of our data and we asked our DBAs, yes, the storage guy asked the DBAs for help on something. I said, what do you guys need for this to not ever happen again, right? And they said, really big, really fast,
06:59
file an object, right? So that's where we got that and then we, you know, kind of kept scaling out from there. Um, so, as we keep going. Here we go. So this is our appliance genealogy. You guys can kind of see on here, right? We started with our one little one and we kind
07:16
of moved forward, and then we had this explosion over here. This was where our business unit needs just started going nuts. We had a lot of new arrays come online here. Um, a fun fact actually about the appliance genealogy that you guys see on Pier One. Um, you're like OG flash blades don't show up in here.
07:36
So we've got our flash blade S over here, right, that we got, you know, last year, but we had one all the way back here that's not on this, and we had a couple of MRAs that aren't on here either from again, back in the day. Uh, so we'll kinda go through that. So this is really what we're gonna be talking about is how we grew and how we handled it.
07:56
So, to set the scene for you guys, when we're at this step one, right, we're, you know, multiplying isolated flashjack systems. You know, I always, I joke with my guys now I'm running a team now. I'm over, you know, architecture and infrastructure. I always when people ask me what does that mean, right, especially non-technical people,
08:12
I say, everything from the power coming in the building to the some of the code and even into the code in some of those use cases. We're the same thing at that, at that time on a much smaller scale, and we had actually the three guys that we had working together, we all knew each other from college and we were kids at that time, right? And, uh, a joke even now, right?
08:32
The kids that I, I'll tell my new employees, right, the kids that set this up had no idea what they were doing, right? And one of the tools that we use to kind of do that is, you know, using Fusion to kind of move through that. But as we grew and grew up and learned a lot of lessons as we went, we started re-architecting and kind of stealing some ideas from how Pure likes to operate and
08:51
bringing that into our, you know, into our environment with how we're structuring all of our data access, right? At first, you know, we had one flash stack. Uh, what's better than one flash stack? Naturally, 3 actually. We went from one flash stack to 3 flash stacks
09:08
and then it came time for our first hardware refresh, um, because unfortunately, not all of our hardware manufacturers have an evergreen model, right? So I had to rip and replace servers. Um, and I got it in my head that I wanted to redefine all of our resources as a true private cloud. Uh, for those of y'all that are like in the
09:25
crowd, y'all know that, uh, that know me, I'm a nerd, right? Badge of honor, not an insult, and I wanted it to meet the N definitions of it, right? One of the things that was really hard with that was dealing with scale, right? There's one of the things in the NS definitions for cloud is to your workloads, it needs to look like bottomless resources.
09:44
That was so hard to get that checkbox. I wish I had fusion at the time because we're using that now for it, and it helps us to hit that much more effectively to work in what, you know, you guys heard this on the stage, right? An enterprise data cloud is how we function with our data now and we started doing that, you know, all the way back in like 2018, 2019, kind of idealizing those principles.
10:06
Uh, then we had that explosion of data protection capacity. I don't know if you guys have heard about ransomware, right? We had to have all kinds of new workloads coming online to make ourselves resilient against ransomware and cyberattacks because, you know, I'm at the Department of Revenue.
10:19
Uh, what are the two certainties in life, right? Death and taxes. So we cannot have downtime in our environment. I think if we, if we have downtime for 3 days, we go into a state of emergency. So that's non-optional. So we had a lot of growth around dealing with
10:34
that and Then we had new business needs coming online, and as much as I tell my boss I am saturated, he keeps giving us new business needs that we have to keep standing up. And I see some guys chuckling because I think y'all have been there before, right? So we're gonna kind of drill into each one of these things and look at the real hard use cases that we were in and how we use fusion to
10:56
make those business outcomes better for us. So the first challenge we're gonna look at is building isolated systems, right? So the flashtack was a godsend for me. It was a way that I could, as somebody with no enterprise IT knowledge, working with some, some other guys I went to college with that had never run anything on
11:13
this scale before, that we could stand up a solution, right? Well, as we started to scale up to more of them, it became really hard, right? We're trying to figure out, 00 crap, you know, I've got this workload configured this way on this array, this one configured this way.
11:28
Where did I put that workload? Where is it at? So how do we solve that, right? So we're using Fusion now to configure a fleet of arrays, and I know I usually don't like the buzzwords that, you know, everybody likes to throw around as platitudes, but the Enterprise Data cloud is one that has been such a lifesaver for us, and fusion is our enabler for that.
11:47
It's how we get to that place where we can truly, truly, truly start treating all of our storage as a disaggregated resource to facilitate our workloads. And one of the terms that I like to use, I've been saying this for a while, so I guess, you know, copyright mic, um, is we call it our capacity platform, right?
12:05
So our storage is a capacity platform to facilitate our business needs because as much as I like technology, when I go to the executive floor and have to pitch for why I need budget for something, I can't say because cool technology, right? I need a real outcome for what it can do for us. So the benefit that we have is, you know, again, that capacity platform has been something that
12:25
has enabled us to keep scaling time after time when these new workloads come. So that kind of brings us to that next challenge that we have is lots of systems to manage this independent units, right? And even more than that is, you know, working with state budget cycles, we have different opportunities that we can grow our storage platform,
12:43
but sometimes we have to shrink it, right? Uh, we've had our MR rays that we didn't renew. We play little tricks. We do flex flash upgrades. I think one of the things, uh, Larry and I were talking about this a couple of months ago, uh, whatever you do when you name your arrays, do not put like the tier level into your array
12:59
name, right? Because we had an X20 array that we upgraded to an X50 array. Yeah, that host name was still X20, right? So we've, we've since remediated that issue. Uh, we have customers with Excel arrays that are named FA 450.
13:17
From uh from like 2014. Yeah, don't, don't array models in your name. Yeah, they uh they're, they're kind of like a bad tattoo, right? Like, let it be a good, let it be artistic, right? Let it be a good thing, right? Because I mean,
13:31
we still, I've got people that, you know, aren't, haven't learned of, you know, pure and kind of what we can do. I tell them, yeah, I've got a raise that have been in that rack for 9 years, and they look at me like I'm nuts, right? They're like, and you're still using it? And I was like, yeah,
13:43
that's like one of the key reasons we like Pure. Uh, but kind of getting back to the challenge that we had, right, is lots of systems to manage as independent units. Well, this becomes a lot easier for us is because we can use fusion as an operational tool too, right? So it's not just something where we can just,
14:01
you know, sling the workloads out there and hope and pray and guess that it's right. It gives us visibility into our entire fleet of arrays, right? Our EDC so that we can manage it more effectively that way. And one of the terms that we use around our shops, so as you guys might be able to see, right, I'm wearing boots, I'm a Mississippi boy, is we call it cowboying or data sometimes,
14:22
right? Because you get a crazy business request that comes in right from the boss's boss's boss saying this needs to be done yesterday, right? So you get in that situation you're like, cool, uh, where does that volume Live, right? So you're opening up all your different arrays, you've copied into your, you know, your clipboard, that you you ID that you're hunting
14:39
for, you're pasting it. Nope, not this one. Nope, not this one. Nope, not this one. Well, this allows us to more effectively cowboy our systems so that we can, you know, deal with these things more dynamically and how we're doing it and not have to look at everything like this individual component.
14:55
So it's not just about workloads using fusion, it's also about managing your total platform as a capacity platform. Um, let's see, going from here. OK. Many different protection requirements, new workloads having to come online rapidly. So I'm not going to go too deep into this one now because we actually have a demo showing
15:16
this. So as we began growing and everybody wanted their applications running on this infrastructure, they all had different needs, right? So we've got our DR systems running on top of VMware. Um, it was Zerto is the DR system we're using it with it. Uh, that one, you know, I don't want to be
15:33
taking snapshots of it all the time because if I'm snapping something that I'm using as a recovery, right, that's, that defeats the purpose of why I'm even keeping that data there, right? It's got a FIFOQ that it's managing for all of its data flow through it, so why would I do that? But you know, it will be nice to keep maybe one snapshot that I roll on there,
15:50
right? So you can have those more, you know, different ways that you can go about it, but then we have our other other workloads like SQL TDE. If you guys have ever had to deal with a TDE sequel before, it's a nightmare. And you have to be so careful because one wrong click can absolutely trounce an array.
16:07
Um, one of the other things that we've kind of started doing through this that I feel like deserves to be mentioned into this is we actually converted everything to NVME over fabric on our systems. Um, so a little anecdotal story, you know, what's a, what's a session with a customer without a couple of stories from the trenches,
16:24
right? Um, so we converted our all of our environments to NVMEOF 100 gig networking on there, and you know, we had some cowboy and we were doing in our DR site because we've been having some, this might ring true with a lot of you guys, we've been having some issues in our VMware environment, right? So our Zerto solution sits on top of that and works with it,
16:44
but because we have all of this power with fusion under the hood where we can move so quickly to doing this, sometimes it's a lot easier for us to just reprovision that DR workload, reseed it, and go. So we were doing that in our DR site. Well, apparently we cleared out a lot of bottlenecks with NVME over fabric because we
17:03
were, we moved some stuff around on the array and then we started hitting a new data seed through a firewall and It overloaded one of our X20 arrays. So it was actually not able to commit IO rights back into the actual hardened storage for the volumes and everything was queuing up in system memory because it was just trying to, you know, compress all that and de-dupe all that data in line going back to the storage.
17:28
Um, I, we almost put it on here. I actually got a text message from my SC on the account and he said, congratulations, you finally were able to overload an X array. But I gotta give it to Pure because the workload actually still stayed up, right? So like, this thing, it can't even fully commit IO to the back.
17:48
Completely unaware of my PMs. Thank goodness, right? Because that would have been a long night trying to explain how our DR works to our PMs and why it shouldn't matter to them. Um, but yeah, so that's one of the things that as we started growing and bringing all these different workloads online, right? Just like we found out when we overloaded that
18:03
array. OK, maybe on these volumes, I need to have some QOS applied on those, right? So we have all these new standards that we have to put on there to protect us, and you know, fool me once, shame on me, fool me twice, shame on you. If I do this again in our DR environments or one of my guys do,
18:20
right, uh, we, we know better than this. So this is a little bit of a bumper that I can put around some of my junior guys, right? So going to the next one, unfamiliar workloads coming online, uh. You know, we work in the, on the infrastructure side, on the storage side.
18:34
You get asked to deploy something you've never heard of before, and you have to figure out what's your best practice for bringing those up. So one of the things that we've actually kind of utilized for this is we've implemented platform engineering on our team. So we have some hiring issues as most government agencies do,
18:51
right? My biggest value add for why people should work for us is because you get to touch a lot of really cool tech, um, and as much as I love my job, I don't do it for the love of taxes, right? I do it for the money and we don't have a lot of that that we can give, but we can use things like integrating our workload deployments into
19:07
a platform engineering practice. So that we can have our younger guys, you know, kind of doing things that they frankly shouldn't be doing otherwise. And I had that same path, right? I'm, I am 2 days on my first big boy IT job deploying a flash stack.
19:21
It's a little bit crazy, right? But we have the tools in place to put safeguards around them to prevent that. And then our more senior guys can work on a strategic level and not get so bogged down in those operational things, right? Because I promise, I have bigger and harder problems for them to work on.
19:37
So presets are a really good way that we can kind of build those known good paths for our workload, so that I have one senior engineer that figures out all the intricacies and complexities of how to set it up, builds a workload preset, and then it distributes across all of our arrays. So no matter where we need to deploy it, not only will it tell my juniors where to deploy it,
19:57
it'll, it'll deploy all of that for them using constructs that are familiar to them, but they don't know intimately. So we hand off to Larry. Yep, I'm going to, so what I'm going to do is go through just a little bit of detail on a couple of the features that Mike is using. This will be quite simple. We have a demo later of presets and then you
20:19
can see demos of more of this functionality if you come by the fusion booth on the show floor. Um, so one of the features Mike alluded to in that sort of use case number one that he has is um unsilo these arrays and right and let and let you be able to manage things through a single UI. What that means to you from a practical sense as a storage administrator is that you have
20:42
some tasks to do, like you need to, uh, you need to add another volume to some host in the environment, but you don't know what array that host is on. Now that all the arrays are linked together with Fusion, you can just log into the GUI of any array, search and find that host, and then configure a volume and attach it to that host.
21:04
And what Fusion will do is remotely send the commands necessary over to the other array to do that for you. So it, it literally changes the UI um of one array into a UI that lets you manage multiple arrays. Um, the same is true for Uh, the same is true for um file services object and object services. Um, you can log into uh flash array running
21:30
file and uh remotely manage file system resources on other flash arrays. You can log into a flash blade, running uh file or object and remotely manage it, manage file or object services on other arrays. So even if you're not using presets, if you turn fusion on, you'll get this. Simple functionality, the GUI, the API and the CLI will all change, and a lot of the purity commands
21:57
have just been extended to work this way, to work across arrays. Sometimes I like to refer to this as the best day in the life of a storage administrator improvement feature that Pure has ever shipped, right, especially if you have lots of arrays in your environment. As a, as a customer, I can actually like agree with him on that. That's not just a pitch.
22:18
Yeah, yeah, uh, OK, this thing is uh only works if you're very close. Um, OK, I got, I did get ahead of myself here. So this is, um, this is kind of showing you how you get to these features in the, in the UI, um, uh, so I have the, the what we call fleet view. So if you're in the GUI and you've turned on Fusion and you've linked the arrays together,
22:39
Um, that little thing that we call a pill will show up that has a local array or fleet view. If you toggle on fleet View, it changes that table to show you those resources all across the fleet. And so if you go to the volumes tab, you'll see that the pods tab you'll see that the file system tab you'll see that if you're on flash blades, you'll see it for all the tabs there as well.
23:03
Um, uh, what we've done to date with Fusion is focus on the more high frequency operations you might be doing. So, um, right now we have that fleet view for all of the storage resources and things that you need to create volumes or file systems, connects them to hosts or applications, and protect them with snapshots or replication. That functionality does not yet work for physical things on the array like changing an
23:27
IP address on a port, but we'll be eventually adding it across all the functionality, and you can see the same features usable on the CLI with this thing that we call the context flag and so you simply use that flag on existing purity commands to tell fusion where you want that command to go or where you want to list volumes from or that kind of thing. Um, OK, back over to Mike. I'll be back for a demo in a bit.
23:56
All right, so, uh, one of the things that we kind of talked about earlier is some of these different workloads that we have with huge disparity amongst them. So we've kind of got, we're going to show you guys some examples that have to do with two polar opposite workloads that we keep online. So the first one that we're going to go through or the first two that we're going to go through
24:13
are SQL TDE workloads and Rosserto workloads. So serial workloads, I need them to be fast, but I don't need them to be ridiculously fast, right? Because it'll help in a DR scenario, but I want my actual applications to win out over my DR, right? Because if we're throttling on storage and my PM comes and asks me what the problem is and I
24:32
say, oh sorry, our infrastructure overran it, I'm gonna get scolded, right? Um, on the other side of that, our applications that our project managers really care about is our SQL TDE workloads, but these come with some problems for us. So TDE is hard to work with because the whole point of TDE is to make everything look like
24:52
random noise, right? No patterns, nothing you can compress, nothing you can de-dupe in there. Um, one of the things that really depends on having patterns and consistency is snapshots. However, there's a really high value that we have for snapshots and what we need in those. So on our SQL TDE, I do want some. I want it to stay for 7 days,
25:12
but my application owners actually flush out their environments once a week and then restore that data back in. That's part of their code management process. But that's really hard because they don't do all of it, so I can't just go, oh, here's a snapshot, right? Do it this way, let me send it over.
25:27
It's a 95% change. That happens with that. So that means I have 95% net new change week over week. So if I have a snapshot, if this 7 becomes an 8, I just, I just 95% increased all of my SQL data stores, which those are some of my heaviest workloads. Um, on the other side of that, my Zero
25:49
workloads. They're FIFOQ. They're continuously changing all the time, and I don't want any real snapshot protections on top of that, right? I'd like to have one as the, the just in case, right? Anybody, uh, in here, when your boss comes to you, right,
26:03
and he's got something going on and he expects you to wave your magic wand and make things better, right? This is for my Zerto workload and it's actually helped us before keeping one snapshot on those workloads because it's uh I think a better, I think a better metaphor for that, right, is for anybody who's watched the Andy Griffith show, you know, Barney Fife keeps his one bullet, right?
26:21
This is my one bullet in my front pocket to go, OK, we're gonna get this one. So we set up the snapshots and we run that. So, well, what about QOS, right? My SQL workloads, I want it relatively open-ended, right? I want that to prioritize over everything.
26:37
Let him eat, right? He's hungry, let him go. This drives better taxpayer outcomes for the taxpayers in the state of Mississippi. As a taxpayer in the state of Mississippi, I am quite appreciative of that. Um, on our er workloads though, it's fine if we go into a sinking state on those and we have to kind of back pressure some of that as long as it gets there eventually because if we've got
26:57
high IO going through our system at a specific time and something crashes, I probably want to go to before that high IO time in my DR strategy, right? Um, so we can set our QOS. These are all parameters that we can configure within all of it, and we can do it dynamically based off the workload.
27:14
So we're going to kind of walk through it and we'll see it where we'll we'll show configuration on the Zerto side of our workloads where it's more restrictive and kind of a, a, a default less, but on our SQL workloads, it's a default more. Um, and on the other side of that is set up our replication. So to fix our TDE problem, we really want snapshot retention.
27:35
So pure released a product for us last year that really helped us with that, the Flash array E, right? So we just replicate it over to the cheap and deep, you know, SS, you know, solid state memory. And we snap it there and we just retain it forever, right?
27:48
Because retaining 100 terabytes of snapshot data on a flash array X is not a great use of taxpayer money, right? And as a contributor to taxpayer money, I take that personally, but a flash array is a really good place for us to get that value out of it. But the problem is is my juniors don't really understand what's going on there,
28:06
right? They're kind of confused about how some of the more advanced topics work, so we can wrap that up into our pre-builts. But on our Zerto, I don't want it to replicate. It is the replication. Why would I replicate that to another array, right? Again, poor use case of taxpayer dollars.
28:21
Um, so with the deployment as well, some of the things that absolutely grind my gears, we all have our pet peeves, right? One of them is config rift. We were on a call actually setting up some of these features. I was kind of talking about some of the things we were doing, um, and this was as we were going into an audit prep with the IRS because yes,
28:39
we also get audited by the IRS and it is terrible. Uh, so I go into my environment and I was like, who set this up? This is wrong. These things are wrong, right? And this is as we're turning fusion up in our workloads. Um, the other side of that is workload configs are baseline,
28:53
right? So we have a standardized config so that when I stand something up, the expected outcome is going to happen because unfortunately, my GPT doesn't exist yet, right? I haven't like integrated my code to have an AI running inside of me. So we're humans and we make mistakes. This helps. to make less mistakes.
29:10
And then on the other side of that is uh workload labels. That's something we use as part of our platform engineering practice to really help understand how we integrate pure outside of Pure's context, right? Because we're not just running storage. We've got a whole lot of other components that my team is responsible for, right?
29:26
Firewalls, networking, all those different things. Those help us to stay on top of that. So I think one of the things too is I, I don't want to go without mentioning this. We set this slide up, right. How do we configure that? You, I wish I, I'm gonna start keeping a
29:39
baseball counter of how many times my juniors walk into my office after I send them a team's message, message to them, asking them to, and I say, I don't know. Did you check the documentation? So. They, they get real tired of that one. but then they go check the documentation and figure it out,
29:56
right? So this helps us to have less of those, how do I do that conversations? Um, cool. So on to the demo. Alright, yep, so we're gonna show you a demo of what Mike's doing, um. Um, also, yeah, I find it interesting that
30:12
you're, you're an employee of the state, but you don't get a, you don't get an employee discount on the time. That was I asked about that in the interview. Uh, I'm, I'm, I'm glad chief legal counsel wasn't there. I might not have got that job. Yeah. All right, so, um, um, uh, one of the, uh,
30:29
one of the things I'll show you in this demo is the two presets that we created for Mike's environment and one of the cool things about these is you can actually upload and download presets in and out of your fleet. So Mike made some of these in his environment. Uh, he downloaded them as a, as a JSON. I took them into.
30:47
My lab and I had to make a few changes because the name of our array that we want to replicate to is different, but I was able to just open the preset that Mike made in my lab, change the name of the target replication array, and then record this demo. So these are, these are the presets for mics and and it helped me not to have another conversation with my CIO.
31:07
I get in trouble when I have those for some reason. OK, um, and so you can see the two, presets here. There's a, a preset that Mike uses to configure the Zerto volumes. There's some properties you can see over there on the right. These are, there's these labels. That Mike wants to have applied when things get
31:25
deployed. This is the preset for deploying the storage for the SQL TDE environment and some of the attributes like what data center we want to label it as being part of and that sort of thing. I think one of the things too on there that you can actually also put controls into some of your presets for your volume sizes in there. So back to that thing about humans make
31:46
mistakes, right? You can actually put maximum size capacity for some of the labels that you're using to apply that. It's fat finger protection, right? I've definitely clicked a 01 time too many and accidentally stood up a 900 terabyte volume, right? That, that could be a it's actually bitten us
32:03
before we overprovisioned a volume for space it didn't have, right? So again, that air protection. OK, so the way you deploy a workload from a preset is by going to the workloads tab, picking the preset for the workload you want to deploy, and then hit continue and you'll get asked a few questions. This is kind of like a form you fill out and
32:23
you guys get to sort of build this form in your environment by the attributes that you put into the preset. So this is the Zerto preset, which doesn't allow much configuration. Here's the thing that Mike was referring to where So this preset only allows you to specify the volume size. There is a maximum size of 100 terabytes, so Mike's junior admins won't be able to make the
32:48
volume bigger than that. And then they pick, uh, they pick the array that they want to place that volume on. Here's where. Fusion will talk to the workload planner in Pier one and get a placement recommendation for which array is best to put this workload on. You can only put this workload on the kind of
33:08
array that you designed in the preset, and then you hit deploy. Fusion will go off and do the work for you to configure the volumes with all the right properties and configuration. Yep, and I think one of the things that's a really good call out on this is because after it stands everything up, it's not creating a bunch of,
33:27
you know, constructs that we're not familiar with, right? You guys see these volumes that it's making on there. These are all things you guys have seen before, right? It's just a lower friction path to how we can get there. And actually, so much so it helped me. One of the demos that they had that they set up
33:40
in there, I swear I hadn't seen the keynote before this. They talked about standing up a new AI workload and doing the whole grafina thing wrapped around it. Um, so we've automated a lot, right? It didn't take 8 hours like he was saying, but it did take me an hour to go ahead and set up the exact same thing, but we were able to go from a full day presentation,
33:58
you know, for how long it would take to stand up a workload like that, and it took me about 1 hour. I started at 5 o'clock and I was on my way home by 6, right? So this is one of those, it helps us to get to those lower friction environments because unfortunately, I don't have, you know, on-prem clawed running where I can kind of automate all
34:13
of those things away. We'll get there eventually, hopefully, fingers crossed. Um, all right, and it will deploy the other workloads. So this is the uh SQL TDE storage for the SQLTDE database. And so what you can see on this, um, on this workload deployment form is that you also,
34:32
you always have to put in a workload name and a workload is simply a new construct that we've introduced in Purity that we wrap around the volumes and the protection groups and the resources that belong to this workload. Um, and, uh, like what Mike was alluding to there is once Fusion has provision to workload from a preset, the volumes, the P groups that you get,
34:55
they're just the standard volumes and P groups that you already know how to manage. You can rename them, you could move them into a pod and replicate them with active cluster. They will still always be part of the workload. Like we won't, we won't lose the fact that they're part of the workload and forget that they were in the workload because you've renamed them. Oh, and I wanted,
35:12
I also wanted to mention that. One of the features Mike is looking most forward to is um later this year, there will be a resource naming policies in these presets, so you'll be able to define how the volumes get named, how the P groups get named, so you can match those to your practices that you that you have. Um, you know, we, we always kind of envisioned, why don't customers just use tags for
35:35
identifying what belongs to what? and Fusion has powerful capabilities for labeling and and forcing uh your users to Um, to do tagging, you can see that here, um, in this, uh, in this demo, um, so in the prior, uh, deployment for Zerto there were specific tags that were forced and not configurable, but in this one,
35:58
Mike set it up so that when workloads are being deployed, you can manage what these label labels or tags say and set them appropriately for the kind of workload you want to deploy. And and that integrates very well into our platform engineering practice. It may all of that can be relevant through those JSON files Larry mentioned earlier. Yep. All right.
36:23
So this, um, this preset, sorry, I got ahead of myself there. Um, so this preset deployed the two volumes for the SQL workload. It set up a protection group for configuring the snapshots with the right retention. It set up a protection group for replicating that that protection group to a third array. And in this case, Mike has a very specific
36:46
array that he wants these workloads replicated to. It's that flash array E. And so when we look at that protection group, you can see that that was hardcoded into the. Preset. You can set it up to be a selectable list, find other arrays in the environment. Let me replicate to the array I want,
37:01
but Mike wants these workloads replicated to a very specific array. He doesn't want his junior admins pointing a massive amount of data at the wrong array, they'll do it too. That's the problem. They'll do it. Yeah, OK, alright, we got a few more slides. I'll try to get through, um, so future expansion plans.
37:24
What's next, right? Um, I've been waiting for, I was told we're going to have a slowdown in Q1 of 2021. I'm still waiting for that slowdown, right? So. One of the things we're doing now is a function not a lot of people are aware of that the Department of Revenue does in Mississippi is we're actually in the liquor business.
37:42
So we are an alcoholic beverage control state. So we operate a warehouse and we import all of the wine and liquor into the state and then sell it to uh private stores and businesses like bars, restaurants, and the sort. Um, so this is such a good opportunity for us.
37:58
One of the things that we've discovered is that Not being hypervisor agile is a business risk now, right? A lot of people kind of got clipped up in that. So this is gonna be our first, you know, kick into having that hypervisor mobility where we have some new systems coming on there. Um, it's such a good, this is literally a once in a career opportunity for us,
38:19
so we're building a new warehouse right now. Um, so we've got, I always, you know, my chief legal counsel loves it when I say this, um, we're building our warehousing space is gonna be about 300,000 square feet, give or take 300,000 square feet, right? So we're building that and I don't know how I convinced my boss to give me purview to
38:39
completely rebuild and re-architect the business process that's driving it underneath that. So we're getting to drive all of those things. And this is where we're introducing some new features. One of the things that's gonna be key for us, and you guys saw this on the keynote stage, is that move into an openshift environment and being able to integrate that into our platform
38:59
engineering practice. But one of the things that scares me on that is I have years, almost a decade because I'm getting old, of experience on the VMware platform, right? That's a lot of hard lessons learned at a smaller scale that I don't have anymore.
39:14
So this allows us to take those, you know, those hard lessons learned, and I can have one guy that goes out there, when I say one guy, it'll probably be me, right? Uh, we go out there, we figure out what we have to do to make our workloads work for us, right? And we can embed those into the presets and now
39:32
as we have to turn up new workloads in this entirely different construct, right, because OpenShift, even OpenShift virtualization is not, you know, your standard virtualization platform. It's containers, right? So it's a different way of operating. So this allows us to build those best practices and those known good pads for how to deploy
39:51
workloads and do that repeatedly at scale as we have new workloads coming online. This new warehouse project that we're standing up, it, it's gonna, it's funny, it's gonna drive me to drinking, right? Um, so with it, I have 19 different IT specific projects associated with it for all the different components we're standing up.
40:11
I, I need to be able to best practice it, know what's good, gotta figure it out. Let me go to the next one. And when I got to come back and do more of that, I have that already worked out and I don't have to double my work up again. So, um, that's on to, that's actually a really good segue.
40:28
Into what our next one is, are moving to platform engineering practices. So we are an extreme use case of how far you can push platform engineering because our team right now is me, a senior guy, and two juniors running all of our infrastructure components. So we highly leverage these practices so that we can move faster and more effectively with our business units because again, back to that thing, right?
40:53
The tech's cool, but my execs don't care about the tech, right? Um, and it also allows us to kind of mitigate some of those config drift issues, and our misconfigurations don't hit as hard because they're, we have bumpers put in place on those so they don't hit us as much, and we can embed those lessons learned into our deployments and into our platform engineering
41:15
practice on a broader scale even so that our juniors can do more things. Um, let's see. So it also helps us with the rapid onboarding of our new workloads, right? We have a lot of them that we're doing, and I feel like I get asked all the time, right, hey, we got to stand up a new workload. I need it yesterday, right?
41:33
or a new environment or something like that, like that one that I did 3 weeks ago using Fusion. This helps us to move more effectively into those things and I can wave my magic wand and my boss goes, Oh, thank you so much. This is why you get a paycheck. Um, the other thing is it lowers the barrier to entry as well for my junior guys.
41:51
I really respect that because I would, that opportunity was given to me, working with the Pure Solutions, and we're pushing that even further because they have more bumpers, I'm letting my junior guys do even more. Um, So now we kind of look at some of our operational benefits of this, right? This kind of is really a good summary of what
42:10
this does for us. So it allows my juniors to do more work, to do more administrative work, and my senior engineers can start thinking more like an architect, right, and not get stuck in the operational. They don't have time to get stuck in the operational components of this. And beyond that, it helps us to reduce the occurrence of misconfigurations,
42:30
not just from an operational perspective, but from a security and compliance perspective as well. MyIO loves that part of it. Um, and the danger to operations. We've all been there, right? I'm not gonna say which one of my guys did this because they're gonna be watching this recording.
42:44
One of my guys shrunk alu once, right? We had snapshot protections in place for that one, so while it did stun the workload, we're able to get back to an operational state in like 5 minutes. Um, and now, my senior guys can focus on strategy. That's where I need them, right? They know our workload.
43:02
They've cried and bled over that workload before, and so that they can use their efforts on higher level outcomes. Get closer. There we go. Very short range on that thing. Uh, OK. All right, so I'm just gonna, um, close us down
43:23
with this last slide. Um, so how do you guys get access to this stuff and like what versions of Purity is it in? Um, uh, I've segmented this by the two main features, so fleets, that's being able to securely connect your arrays together, and being able to forward those commands remotely between arrays,
43:40
have fleet view and the GUI, that kind of thing. We started shipping that in Purity 6.8 on Flash array and for flash blade. Purity 4.5.6 and that's available for all the protocols, uh, um, block file and object across those two arrays. A Clockto is also supported. You do need to be on these hardware models um
44:02
to get support for it. The simplest way I can say it is that you have to be on an X3, sorry, an R3 or newer array. Uh, you cannot run Purity 6.8 on the X2 arrays. Uh, they only run up to Purity 6.7, so you have to get on an R3 or new array to run it. Um, the other key requirement is that Fusion does not provide an identity
44:26
provider. We rely on Active Director or LDAP to be a cross array identity provider. So if you want those remote management commands to work when you log in, you have to be logged in as an LDAP user. So you have to configure all your arrays to connect to an 80 or LDAP server and authenticate that way.
44:45
Um, and obviously the rays have to be able to reach each other between the management ports. Uh, the presets and workloads that that Mike is using, we've shipped those for block workloads on Flash array, and on cloud Blockstore. They're going to be coming, um, the end of this calendar year for file workloads across FA and FB. And then you heard in the keynote this morning
45:05
we're. the object protocol to Flash array and then we'll be adding uh uh object support in presets, then that'll work across Flash ray and Flash Blade as well. So the whole idea here is that we get you out of the business of worrying about all the little tiny knobs and get you into the business of uh setting a standard and repeatedly deploying things according to that.
45:27
Standard. You can do this like in Mike's environment, there's not a lot of heavy API use and you do some, but the rest of your team doesn't. You can do this with the CLI and the GUI. You can also do it with an API and there's another session this uh Thursday that talks about these same features but how you can drive them with an API and uh with like answerable and that sort of thing.
45:50
OK, I think I need to put up these QR codes so you guys can get some cool socks. Um, but that's all we had. Thanks for listening. It's great. I appreciate it guys. Thank y'all for coming out.
