00:03
So we'll just jump right into it. Get started. We're gonna go back and forth a bunch between me and Brent. Um We want to try to save questions till the end. I think we'll probably have some stragglers coming in through the session. Maybe they won't hear that and might throw questions are up in the middle.
00:17
Uh But we'll see how it goes. OK? So the best way to understand what fusion is is to think about the actual definition of the word fusion and this is why we chose this word to represent what this product is doing. So fusion fuses your rays together, right? We're going for the same spirit of the definition as the uh what the word fusion means and you smash a bunch of things together.
00:42
The resulting reaction that you get is a lot more power and a lot more outcomes out of having those things together. Specifically in this session. We're gonna talk about how the new version of fusion that we're shipping later this year is gonna let you transform the existing infrastructure that you have into a multi array storage platform to give you some more benefits,
01:03
those benefits. Are listed here on the side. I don't want to read them word for word for you. But what the, the values that we're going for is to increase your or operational efficiency. And this does not mean forcing you to automate everything with code. That's a part of what fusion can do. And if you automate things with code,
01:21
fusion is even going to make it easier to automate things. But with fusion, we're gonna drive operational efficiencies, whether you're using a gooey, a cli doing things manually or automating things with code. Um We're gonna miti mitigate configuration risk. This is um fusion helping you to make sure you're configuring storage the way you want, the way your uh corporate standards are defined
01:45
and make sure things like data protection get turned on and that sort of thing. Um We'll also be doing some things in the future to improve workload visibility and to uh allow you to increase utilization by moving workloads around. And then um you'll see in the demo how we use the A I workload planner from the pure one cloud to help you do workload placements um to figure out where to best put workloads in the
02:10
environment. OK? Um So fusion is a control plane but it's not like a typical control plane. It's not something extra that you have to download and install and configure and maintain in your environment. Fusion is now actually embedded into the purity operating system.
02:29
It's just part of purity. It's right there in the flash array and flash blade rest api so you can think of fusion as new capabilities and features that you get access to just by simply upgrading the version of purity that you're running on your arrays. And there's no license cost for fusion. There's no cost to use it.
02:48
It's not a separate subscription, there's no need for you to purchase any uh access to it or anything like that you get on the version of purity that this, that these features are available in and then you can start using them just like every other pure storage feature has been now. Um a, a pretty interesting thing that we do with fusion is we're distributing the control plane around all of the arrays in the
03:12
environment. So all the arrays that you connect and link together to make up your fusion fleet, we'll have a distributed control plane where the arrays can talk to each other and share configuration information. There's a couple of new features that we're adding later this year that are gonna make use of that distributed uh config info and Brent's gonna talk later in the session about how those
03:32
work in the back end and what that means for you. And so that's what we mean by saying that the uh critical capabilities like the new enhanced provisioning work flows that fusion is gonna deliver are distributed around the environment. So you don't need access to a particular point in the control plane. In order to do things, you just need access to the array that you want to do a thing on or to
03:54
the arrays that are gonna be involved in the things you wanna do, like creating storage and setting up replication, you can drive that from any array in the environment. Um Now when it comes to connecting your arrays together, what fusion is doing is linking or connecting arrays together on the management network.
04:12
So one of the requirements to use fusion is that the arrays that you want to link together have to be able to reach each other on the management network. They don't have to be in the same subnet, they just have to be able to route and reach and access each other. That's one requirement. Another requirement to use this functionality is that we are relying on a central ID P
04:34
like uh active directory or LDA in order to drive authentication through uh through the environment, like we don't want to change the security posture of your existing environments. So if you want to use fusion's remote storage provisioning capabilities, the users that drive those features have to be authenticated by AD or LDA, which is what authorizes them to be an array, admin or a storage admin on multiple arrays,
05:02
both the array that's initiating the command and the other array that's receiving the command. So those are the two major requirements you have to worry about, they have to be reachable across the management network and you need to have uh ad and LDA configured with the same uh group, the same uh active directory grouping information defined on each array that you want to be in the same fleet.
05:25
OK. So one of the things that you hear us saying over and over again and Larry has mentioned it and I'm gonna say it a few more times is now that fusion is part of purity, any array can manage every other array in the fleet. This is because arrays can now communicate and coordinate with each other. We call this array federation.
05:44
This enables things like remote storage management and you can actually manage all of your arrays from a single array. Is this so to get started with a a federation, we first need to create a fleet. You can create a fleet from your existing array. We're gonna be showing you a demo of how to do so shortly.
06:04
But the important takeaway is that this is a completely non disruptive operation. So no need for get it down time, no need for workload migration. Just a few simple clicks and your arrays will be Federated together and ready to be managed as a fleet. So the first thing you would do is pick any array and create a fleet. This will create a fleet of one,
06:24
then you will go to another array and add that array to your existing fleet. This were non disruptively, add that array into the fleet to give you a fleet of two. As part of this process, the arrays would do a secure handshake exchange certificate so that they learn to trust each other. Arrays would only talk to other arrays in the same fleet as part of the security requirement.
06:47
And then we'll just go to the third array and add that to the fleet of two to give us a fleet of three. I think you're getting the idea. So now that we have a feed, what are some of the cool things we can do on the fleet? For starters, you can point your browser to any of the arrays existing Gui. There's no need for a special separate, specialized Gui to manage your feet because any
07:09
of the existing arrays Gui can now be used to manage the entire fleet. This is true also for side radio fermentation like so basically, if they have an international presence 20 what are the requirements, the rate? So I'll be covering this in more detail later in my talk.
07:40
But the short answer is there's no latency requirement, the fleet can spend geography. So as I was saying, any array can be used to manage the entire entire fleet. There's no need for a specialized separate go. The same is true also for C and A P and so any storage management C go and A P now not just work locally but they also work in the fleet. So for example, if you are a storage admin and maybe you got a request from a DBA saying,
08:11
hey, they need more storage for cluster three before you would have to first figure out which array was physically connected to cluster three. Maybe you wrote that down in a spreadsheet somewhere or maybe you just have to go into each array to figure out is this array connected to cluster three. Now you don't have to do any of those. You can simply go to any era,
08:29
go and say, hey, I want to create a volume for cluster three. Doesn't matter that the array that's actually connected to cluster tree is the third array. You can still create the volume on cluster tree while connected to array one. Another example, maybe an application owner wants to resize their lung and all they have is the volume serial number.
08:51
Again, previously, you would have to hunt down this volume going array to array seeing which actually has that volume. But now you can just go to any array school. We go to the familiar volumes page, look up the volume by serial number and resize that line even if the volume is on a different array from the good that you are connected to
09:19
see as you see structure. How can I recharge myself? So that's a great question. Again, I would cover this later. But the short answer is that you can refer to a volume by its name or also by its array prefix. So you know exactly that this volume is on this
09:47
array. So let's just go straight into the demo. We're gonna show you how we can take existing erase, federate them together and start managing them together as a fleet. In this demo, we have three arrays, flash ray 12 and three. And you can see this is using the existing array, but we have added a new option called the fleet where you can create a new fleet or
10:11
you can join an existing fleet. So let's create a new fleet, given a name, I'd say my fleet and click, create a fleet. That's all there is to it. Now you have a fleet of one. We now want to add the remaining two arrays into the same fleet. So let's create a fleet key copy to the
10:25
clipboard and now head over to flash array to school. Same thing. We head over to the fleet tab. This time, we want to join an existing fleet. So we enter the fleet name and we paste in the fleet key and we hit join the fleet and boom. Now we have a fleet of 21 and two,
10:45
just connected together all of the data seamlessly imported into the fleet. And we now want to add the third array into the fleet. So same thing will create a feed key, copy it and then go over to the flash array street. Go click on the feed tab. Do you know the drill by now?
11:02
Join an existing fleet, enter the feed name and give it the fleet key. So now you have a fleet of all three arrays ready to go. So what can we do with this feed? So we can go to the storage page into the volumes page. And we have introduced a new option here called feed view. So here you can see all the volumes on all the
11:23
arrays. So volume of flash array, one volume of flash rate two and volume of flash rate tree. You can have volumes of the same name on different rays. It doesn't matter because you can disambiguate by which array is actually physically on. You can also see the same thing for host and host group here,
11:37
the host on flash A one and two and host a tree and here's the host on flash A 12 and three. So if you remember the previous example where like the application owner wanted to resize the line and all they had to go by was the volume zero number. Well, let's just take the volume zero number that they have given us. Copy it, go over to the VOLS page, hit details
12:05
and just paste it into the search bar here. Hit enter and boom. You found the volume. You see which volume corresponds to this, your number. You see this on flash area two. And you see that its name sequel B 1001. So now you can click on it to drill down. You can see there's 10 gigabyte and you can see that this volume is actually on flash to
12:23
remember that we're still on flash array three school, but we're doing all of this on a remote array. So now we want to resize the volume, maybe we want to upgrade it to a 20 gigabyte volume instead of a 10 gigabyte volume. And there you have it just all the familiar things that you used to done locally. Now you can even do remotely and now we want to provision additional volumes for the same host
12:44
group. So you can go to the usual create volume page specify, you want it a flash ray to specify the name and size, look for the volume and connect it to the host group that we were connected to the previous volume. Everything done on flash array tree go you no longer have to switch tabs back and forth.
13:08
You can just manage your entire feet using the same array. OK? So what Brent just showed you is see, Google slides is a not a very trustworthy way of driving sessions, I guess let's try to fix. OK. All right. So uh so what Brent just showed you was
13:28
basically some of the simple infrastructure and underpinnings that we had to build inside purity to enable much more powerful workflows. And that's a demo that I'm gonna take you through a bunch of built in automation things that we've done. So what you just saw array federation and remote storage management, the ability to have a command initiated from array one and sent to array three to take
13:52
action on storage objects over there. That is being, that's the first fusion functionality that will get introduced um later this year. It's coming for flash array as part of the 68 release line. That's when that functionality will start to show up that what you just saw, Brent showed you remote volume provisioning remote host
14:11
connection, connecting the volume to a host remotely, you can also create host remotely. Um This was an early build of fusion when we G A this functionality that you just saw on the demo in the 68 release line, that functionality is gonna apply to everything, protection groups, pods, hosts volume groups, everything you can do in the gooey, those commands can be forwarded between arrays. So you can just completely manage things from
14:36
within one interface. Yeah. So the question was, is it gonna include system configuration stuff? Um Not yet, but we're thinking about doing that in the future. So right now, we're focused on the storage provisioning tasks of getting uh volumes and then later file systems and buckets provisioned and connected to the host or applications that
14:58
need them. Um But we are talking about in the future of um um spanning into the actual system, config stuff like DNS settings, network interface addresses and that sort of thing. That was a pretty good question. OK. So what you're gonna see in the next section required us to build all that preliminary architecture to allow a ray to securely
15:18
communicate and forward commands together. And this is what we call um a built in automation or the fusion automation engine thing that now lives inside of, of every puree um starting with that 68 release. Um So I'd like to use this analogy um for what we're doing with this new feature. And um if any of you are old enough to remember from before, uh fast food restaurants had combo meals.
15:44
Um Just about every storage administration team I talked to remembers when uh combo meals were not a thing, but I did talk to a team one time that had a pretty young guy and he, he actually said combo meals weren't always a thing. Um um So it's interesting to use this analogy and see uh who it connects with or who, who it doesn't. But if you, um if you remember for me,
16:06
I remember riding in a car uh as a kid with my parents and they go through the drive through and me and my brother would be firing uh all the details about what we wanted at our parents and they're trying to repeat it into the drive through uh system and you had to um, discreetly say everything you wanted, right? You had to say every detail about you what you wanted and everything in your meal, things would get screwed up,
16:30
things get done wrong and things will get translated wrong between you and whoever was making the order, things would get done wrong back in the kitchen, in the restaurant. And so you had lots of options. There was lots of things to customize. It led to more errors and a poor experience. And the industry addressed this by introducing this concept of a combo meal.
16:52
And that did things for both the provider, the or the restaurant who provided uh the food and, but it also did things for the customer or the consumer. And what it did for the provider is it allowed them to standardize on a few specific configurations and get really good at delivering those quickly and with consistency and for the consumer, it let them quickly state what they want,
17:17
but still make a few customization options. And soon now when you go to a restaurant, you can just say I want the combo number three and I want uh this for my side and I want a small drink, right? And if you're translating that to your significant other, because they're on their way home from work or something and they're gonna pick up food, they don't even have to know what a number three is.
17:37
You can just say, give me the number three thing. And so this means there's less mistakes. This means you get a faster experience as a consumer and the provider. In the case of fusion, the storage administration team gets more standardized conf figs and a simpler provisioning experience.
17:54
Now, if you translate this into the way the storage management world works, if you're the storage provider or the storage administrator, you have a lot of things to worry about. Where are you gonna place new workloads? If you've got a lot of arrays in your environment, what's the configuration requirements that you're after? These can be configuration requirements that
18:13
you've defined in your environment standards that you set up. If I'm gonna run an oracle database for this particular part of our business, it has to have this kind of snapshot retention, it has to have safe mode, it has to be replicated across different availability zones. These are the kind of things that fusion wants to help address today.
18:31
All of this stuff is stored in external run books, spreadsheets, maybe scripts that you might have written integrations you've tried to build and that sort of thing we're trying to get this, uh We're trying to get the storage platform to help you solve these problems uh directly within the system. And so what fusion is essentially doing for this use case is becoming a layer that's it's
18:53
between the, the the the people who are provisioning the storage and the back end systems in your environment. And this can span across flash blade, flash array, cloud box store in the cloud block file and object in the future as well. There's a bunch of um attributes we after delivering with fusion listed across the bottom there, what you're going to see in the demo coming up is workload,
19:16
placement, risk mitigation, um uh config synchronization, data protection, observ ability and some of the other things we're, we're targeting for the future as well. Now, I'll take you through a demo of this. These demos are all pre-recorded. So we've had to become very practiced at talking to them while they uh well, they play here. Um OK.
19:39
All right. So here, um we've already got a fleet and um we will be introducing a concept called availability zones and you can organize your arrays into a Zs and regions. Some of these arrays are organized that way. Some are not in an A Z uh yet. And we're introducing new workflows in familiar places. So normally you would go here to make a new
19:57
volume, but now you can do this thing called deploy a workload and deploying a workload means that you get help with provisioning more than just getting volumes created. And we do that with this thing called a storage preset. You get to build and create these here. We're gonna use this one that we've set up for oracle block workloads. You give your workload a name,
20:19
you select the storage class, which is really just selecting. What kind of array do I want to run this workload on? Um Here we're gonna pick flash A X. Now the workload properties that you can set up define the workload type, the kind of replication you want. In this case, near sync. Where do you want it to be available across uh
20:37
zones? Do you want safe mode enabled? And then you can apply quality of service limits to it like IOPS and bandwidth limits. Another powerful feature is that you can inject your own custom labels into the process. Here, we're going to make one called a billing id. You can create multis select labels where the person provisioning has to pick uh some
20:57
attribute about the workload. Here, we're gonna have one called tier and pick mission critical. Now, this preset is gonna create for us data volumes and log volumes of different sizes and quantities with different snapshot configurations and retention. Um This is all customizable. You can set the presets up the way you want
21:16
with lots of different kinds of volumes. Now that the eraser and a fleet, we can find a host or a host group anywhere in the fleet and designate. That is where we want this application to be connected. Um We can pick an availability zone that we want to deploy this in and then we can ask pure one to give us a placement recommendation to recommend which array is best to put this
21:39
workload on based on all those workload attributes. And we pick the array that we want and then we move on to the next uh step which is configuring the target array. Here, you can pick a storage class for that as well. We're gonna pick Flasher ac we want to replicate it to a different kind of array and a
22:01
replication target. Now we have that zone constraint. So this is going to mitigate the risk of someone provisioning replication in the wrong place or to the wrong array or to the wrong zone. And then you can review all the configurations that you put into the preset and then ask fusion to deploy that workload.
22:18
Um Now with the new fleet view capabilities, you can quickly toggle over and say I'd like to view those new things that we've provisioned. Now, fusion did all of this for you. We went to the Gooey of array one, we got volumes provisioned over on array five. We're getting near sync replication what we call active Dr set up for us to array eight.
22:39
We're getting snapshot schedules enabled Q OS limits, applied labels, put on all the volumes and, and storage objects that are gonna get provision for this. It's getting connected to the host that we wanted. Um We didn't have to bounce around, we got to go to one array, go to one place in the Gooey, do a simple workflow tell fusion.
22:56
We want all this work to happen and fusion goes off in the background for us and does everything. Uh One last thing I'll mention before I turn it over to Brent to sort of deep dive into how some of the back end pieces uh work is that um we, we've made, we've made fusion do this to simplify the overall provisioning process. And the demo, you just saw applied to a block workload on a flash array,
23:20
we'll be doing this uh the same thing for file workloads and object workloads. And depending on the kind of uh workload preset you want to define, you might have different properties showing up in that preset that you build. So if it's like a file preset, there could be audit properties. Um There could be uh access lists. Um uh If it's a bucket work uh uh workload,
23:44
there could be a replication for that. Um When you look at the flash array and flash blade systems, they're both very similar to manage but different capabilities. Uh uh In the two products have different sort of knobs because they have different features. What we're going for here is that, that storage preset gives you a very simple and consistent
24:04
declarative way to say I want snapshots, I want protection. I want these labels and these stuff applied and you can make a preset for any kind of workload and you get a consistent uh outcome and you don't have to go do all these discrete steps yourself. All right, you've seen the demo, you know how it looks, let's do a technical deep dive to see how it actually works.
24:28
So the main thing that Larry short was that you can now deploy workloads using this thing called a preset. And as he mentioned, preset is really a predefined template that specifies how workload should be configured. He also mentioned that you can now create presets. And once that's created every array on the fleet actually gets access to the same preset.
24:49
So you don't have to create the same preset over and over again, you just have to create it once and it's available to all arrays in the fleet. And when you add a new array to the fleet, the the new array also gets access to the presets. And more importantly, when you make changes to the preset, every array sees the new changes. So you only have to make the change once and
25:09
how that works is with this thing called feet synchronization. It's built on top of array federation. So now that arrays can talk to each other, they can also synchronize the objects of each other. So in reality, every array actually just maintains a local copy of the presets but kept in sync with every other erase copy.
25:30
So this gives the view that there is just one copy of the preset. But every array actually maintains its own copy. And this is also how change propagates throughout the fleet. And this is an important design element because the array being able to maintain its own local copy means that when it's disconnected from the other arrays can still provision workload with its own copy of the preset.
25:53
And this allows the architecture to be scalable and for tolerant. So to that gentleman's question over there, this is one of the reasons why our feet can actually span geography. Mhm As as you do with larger and larger parts and it cannot some kind of multi tendency or some kind of access allow some people to see some presets
26:22
and edit them or just read them. Um So great question. The answer here is a bit more forward looking. But the idea the thinking is that there's going to be more than one way to like group your array together in the feed is more than just all of the area that shows up in the feed. Larry has already shown you, you can actually group your array into a and regions.
26:45
So you might envision additional roles that being created like more than just a free admin. You might have like a zone admin that can manage the resources and the policies that this zone is using. So you could possibly like the market it like OK, here's my emir data center, I'll create an easy for it and then I'll have the zone admin just manage this data center and then I might have like an,
27:04
a data center and for things that are shared across all my data centers, I could have feed white policies that just applies to all of them. So, laws. That's right. Yep. So, but right now you can already have a race across geography. You can even have like cloud block store that's
27:22
running in the cloud as part of the fleet. And in fact, it doesn't have to be just flash arrays. You can have flash blade, you can have flash arrays, you can have cloud block store all in the same fleet. That's because go ahead, right or both. So we now have flash blade, flash and club
27:47
block store that can coexist in the same fleet because fusion is now part of purity and purity runs on all of them. So they all just benefit. But that doesn't mean that we support everything on every array. Now, each array flash flash array club block store still maintains the unique capabilities. Like Larry mentioned flash bed is still our scale out a platform that supports object and
28:08
foul flash array is still scale up architecture platform that supports a file and block and club block store supports block. But what we have done is provided a consistent management experience to manage across your flash array, flash bit and club block store. So with presets, you could have presets for like object workload, you could have presets for F workloads or you
28:30
could have presets for block workload. And based on what presets you have chosen, we can give you suggestions, we can direct you to the appropriate storage asset for you to provision your workload on. And how does that work? Well, we've integrated with pure ones A I driven workload placement engine.
28:47
So based on the presets that you have chosen, it, it considers the capability that's required, it considers the load and capacity projections of the era and it gives you a list of recommended arrays. And from that list, you can now decide which area you want to provision a workload, click on deploy workload and this would create what we call a workload specification.
29:08
This is a very important piece of meta data and it is essentially a yam file that precisely describes how your workload should be configured from the data protection that is required all the way down to what labors should be applied to each resource. Once that workload specification has been generated, the fusion automation engine, which is another thing that fusion is adding to purity would look at the workload spec and it
29:33
would create and configure the workload according to spec. So this workload specification is important for several reasons. As Larry mentioned, we want presets to be consistent across object file and block across flash and flash play, which means we want a consistent way to be able to define like data protection cos replication, you name it workload configuration.
29:56
So for example, at the preset level, you might only have one way to specify a snapshot configuration like frequency and retention. But when it comes time to apply it to a block workload or a foul workload, the fusion automation engine might realize that OK. For block, you need to configure P groups. For foul systems,
30:17
you need to configure file system policies. But what the fusion automation engine would do is you configure all of this automatically on behalf of the user. So the user doesn't have to configure P goods manually or file system policies manually. They just provided a consistent way to provide snapshot configuration information at the same time being able to support whatever we
30:39
currently have is essential to not disrupting existing work flow. So for example, being able to continue supporting P group is critical because we don't want to break existing third party applications of backup applications that depend on it. So worker specification really gives us the best of both words, it allows us to provide a consistent management experience while fully maintaining backward
31:00
compatibility. And the other reason workload specification is important is because we want to do more than just simplify and standardize the provisioning of your workload. We also want to simplify the end to end process of workload management. So imagine you already have your workload provision and now you want to make some changes
31:20
to your workload configuration, say you wanted to increase the retention of your snapshots. What you can do is you can update the workload, spec directly. The fusion automation engine is continuously monitoring the workload, spec and comparing it against the current configuration when it detects the change to just go ahead and reconfigure the workload and increase the retention of your workload.
31:41
Now, this makes it really easy to add new volumes or new resources to your existing workload. Because what you can do is you can just create a new volume and say, hey, this needs to go into my existing workload. This updates the workload spec and the fusion engine will take care of the rest. OK. Uh Something I um I forgot to mention earlier
32:01
was that um um these new fusion workflows. So the storage preset and that preset based provisioning, it's protocol agnostic. So it doesn't matter if you want to connect uh that workload using fiber channel or Ice Cuzzy or NVE over rocky or NVU or fiber channel or NVME TCP. Um R rays support all those protocols and um everything you provision with fusion can be
32:26
connected that way. So um it's, it's um it supports all of the uh block storage and all of the connectivity protocols. Um The storage preset based functionality will be coming later near the end of this uh calendar year. That's our current plan. So the first thing that ships is that array federation and remote provisioning
32:45
functionality. And then later this year, the storage presets. And this built in automation. Now, when you think about the use case types for this, um we encounter a lot of different customers on a, on a spectrum of uh no automation up to heavy automating everything. And we really wanted to design something that would give value to every kind of customer.
33:07
So if you're doing things um manually, meaning some when in your environment says, hey, I need another set of uh storage objects, provision for an oracle database. Um If that's a frequent thing that happens, you can create presets to help you automate and drive that configuration. I also spoke to a customer that um wants to use the storage presets to predefined what their level two storage operations team should do.
33:33
So architects can define the scope of the presets, all the various kinds of presets you wanna have in your environment, that sort of becomes your service catalog of the kinds of storage configurations that you wanna support and allow in the environment. And then you can have a lower skilled like level two storage operations team that maybe reads instructions from a help desk ticket or something to go provision uh storage for an
33:57
oracle database using the the gold preset with synchronous replication or however, you decided to name it. Um It means that they're gonna screw up less things. The conf figs are gonna be more standard and simpler for them to do. It means that um uh that one customer I spoke to that level two team would still always need
34:16
to escalate to the storage engineering team whenever a ray based replication was involved, because the level two team wasn't sophisticated enough to set that up, the presets will automate configuring and doing that. So we're taking all this work off the storage team, taking all this work off the level two team making their jobs more consistent and predictable.
34:34
And you can do that with manual use of the gooey or the cli or if you're a very big customer who's into um automating things, you can automate everything you saw in the demos uh via API S. So API S are being modified to support remote operations. You can connect to array one and forward your API request to any other array in the fleet. You can create presets via API provision from
35:02
presets via API. Everything is still strictly automat. What that means is there's a lot less overhead for you to build in your own automation and your own um tooling. So rather than you having to learn each individual discrete command, how do I do all this stuff? Um You can just provision presets in the environment and then simply and quickly
35:21
provision from the preset. So there's less tech debt in your code. So the idea of preset is really awesome because what do we do for customers that are using the same largely set? And we want to architecture into prees is to create a reset phone where you say add a bottom of ac on.
35:49
Yeah. Yeah. Your question was um how would a customer who's already got all their volumes provisioned and wants to be able to take advantage of some of these new functionalities? How do they get into that world? Um So we will have a pure workload, create command that allows you to put that name into the system Like you saw in the demo, uh We created a workload called or a DB one from this
36:11
preset. So you would have to type the pure workload, create command to, to like reserve that name in the system or a DB one. And then you can um simply add volumes and say now these volumes are part of this or a DB one. You don't need to have a bunch of complex stuff and a preset to do that in the future. We want make fusion monitor the configuration of existing workloads against what's defined in
36:34
the presets. Um We sort of uh we're referring to that as um compliance and remediation, which is a feature for our future, but we definitely want to let customers on board their existing workloads into that paradigm. Other things you'll get by using the new workload concept is aggregated performance and capacity metrics for observ ability.
36:54
If you've labeled things via the preset method, if you've labeled things with like a billing idea, department ID or anything you've made up production, you could put a tag on stuff, you could say how much capacity is that billing ID using? How many iops is that billing ID using across multiple different workloads? Now. Um um So I like to think of this as automation.
37:16
You don't have to automate. So it doesn't matter if you're using the Gooey, the CLI API or SDKS. Um You can simply create storage presets, provision from those. You don't have to build any automation that's going to do things like pick which array to put stuff on um improve the observable.
37:36
You get A I ops assistance from uh from the pure one cloud of where to put workloads. Um I do want to make it clear that everything fusion is doing is happening on prem it's inside the pure rays. The communication is um horizontal from array to array. The one service that we use from the cloud is the pure one workload planner. And that's where the A I workload placement
38:00
recommendations come from. That only requires a phone home connection to get access to that. It does not require the full um two way pure one edge service. So if your rays already communicating via phone home, it will be able to reach the workload planner and inject that placement recommendation right into the provisioning process.
38:19
Um What this means for uh for those of you that are automating your entire environment um with code. Uh It means you get to do all these things on the left with just one simple line of code. So I asked one of our solution architects to write for me the simplest code that they could, they would provision. Uh It's basically doing what you saw in the demo.
38:43
Uh It's doing about 80% of what you saw in the demo. So the code on the left there is creating a number of volumes with different sizes with different uh snapshot schedules and replication label, putting some labels on them uh to define some business attributes, putting Q Os limits on them, uh getting them connected to uh hosts and that sort of thing.
39:06
Um So rather than you having to learn how to write all this code and define all this code, you could simply create a storage preset, which if you want to do, you just go into the gooey and do that. And then you could give someone an API key to be able to provision from an array and they can provision from that preset and get all these outcomes without you having to learn how to drive all that automation.
39:26
That's the main uh value add and operational savings that we're trying to get at with fusion this year. Is there a question we back? Yeah. So with this command, um when you ran through the Gooey, even after you selected the reset, you still went through a process of taking the rays patient or does this, since I don't see any kind of Q and A after
39:49
assume the top level. Yeah, this is a bit, this is a bit simplified. So the way it would actually work if you um so this wouldn't be doing a, a placement recommendation um That that particular command would put the workload on the array that you happen to connect to. If you want to get a placement recommendation via CLI or um or API.
40:13
Um We're making that a two step process. So you would submit an API request once that says I want a provisional workload like an oracle database from this preset. Tell me um give me my stack ranked list of recommended arrays. It responds and then you parse out the one that you want and submit it back and say do that thing. But on, on this array to take any plans that
40:36
workload plan or knowing where I want to put it. Yep. Yeah. Yep. OK. This is the last slide we had. So um um Brent and I um have to run off to another session soon, but um we'll be around in the, in the demo area for questions. Um I also want to mention that um we've made uh the preliminary features that you saw
40:58
Brent demo. We've made that available in pure test drive. Um It's invite only. So if anyone does want access to that and want to get access to a test drive environment and um try that out. Um You could reach out to us um through your account team and we can get you access to test drive to be able to do that. And that's our plan going forward is to put
41:18
more and more of these features into test drives so that you can experience them and try them out. Um If you don't know what test drive is, it's 100% virtual environment that you just connect to a website and it provisions a bunch of uh arrays for you and lets you work through a guide to figure out and learn how new features work. Um So that's all we had.
41:36
Thanks for your time and attention, everyone.
