31:49 Webinar
FlashArray Meets Nutanix: Independently Scale Compute and Storage
This webinar shows infrastructure and storage admins how pairing Everpure FlashArray with the Nutanix Cloud Platform over NVMe/TCP breaks that dependency, delivering external, high-performance block storage to Acropolis Hypervisor (AHV) clusters.
This webinar first aired on 24 September 2026
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
Good afternoon, everyone. Here is Eugenio Grosso from, Everpure Data. I'm a principal field solution architect with Everpure, and today's session, is, meant to lead you through the integration between, Pure-- Everpure and Nutanix.
Before starting, please, mind about the, our, Pure//Accelerate world roadshow. I want to mention that, you can find one, in a city near to you. Just look at the, QR code, the, the, the links that you find in this page and, by the next week there will be anyhow the, our, Accelerate EMEA, in,
London, on next, Tuesday and, Wednesday. As mentioned, the, this, technical session is meant for giving you an overview what the integration with the, Everpure FlashArray and Nutanix looks like. I'm going to introduce, briefly what the in, architecture looks like.
What are the requirements, from both, Nutanix and Everpure for allocating, integrating the external storage into a Nutanix AHV environment? What are the basic steps for configuring, the FlashArray and, through the live demo I'll, I'll, drive you to these, the first steps of configuring the array, which as you will see are, very easy, simple as usu- usual with Pure.
Plus I'll give you, the, restrictions, and the consideration guardrails about the, limitations that there are, within the architecture, and, what is, the, next steps that you can take if you wanted to try to turn this, demo into a practical, test of, the integration and eventually turn, your existing or your, projected, Nutanix AHV environment, supported, feeded, feeded,
feeded by, a Pure Storage FlashArray. The Nutanix Pure integration has been, announced at the, and made GA, the end of the past year. And the reason for having, this kind of, external, storage integration within Nutanix, is mainly due to the fact that, high performance
workload, specifically HPC and AI workloads, they need external storage for, being, u- for being, deemed as first class citizens. So the FlashArray density and the, extreme, high performance IO that, can provide to the, hypervisor environment are such that, Nutanix, AHV environment can,
ben-benefit from that. Additional, features that are, increasing the, the, utilization for this kind of model, are the granular storage control, so every single, virtual machine disk, is mapped into an external storage array, volume so that, it is, every kind of operation that can be accomplished through the array is possible,
with the approach in this model. The, fur-furthermore, there is, preservation of the investment if, either, Pure customers want to move to Nutanix without losing their, existing, FlashArray//ST storage. So, essentially this kind of integration for Everpure, customers is such that
they can, extend their, existing estate also to, Nutanix AHV. Day one and day two operations, are normally, happening through Prisma to, Prisma, either Prisma Element or, mainly through Prisma Central. The, underneath, configuration of the NVMe TCP path is automatically, taken over by, Prisma itself.
So the installation procedure, it, that is driven automatically internally to Prisma, essentially, creates all what, what is needed on the array side, but a minimal configuration that is needed on the array just to allocate the portion of the array that, Nutanix AHV needs. So, in addition to that, that there is the, security posture in, with the embedded,
encryption and immutable snapshots that the array per se provides. Furthermore, there is, jointly validated design with the Cisco, so Nutanix, Everpure and Cisco have this CVD, which covers completely, the ref- reference architecture, of the three, vendors. Quickly through the architecture, the, this is common to any kind of external storage that
has been added by Nutanix to their ma- to their, AHV environment. So, the, connectivity happens through NVMe/TCP. Best practice and recommendation, will, vary according to, sing, cu-customer by customer. So for this kind of, activity, you have to recur to your Nutanix, rep, support representatives to, properly, si- to properly size and, and, and tune,
the, network part, to, fulfill the requirements for, the, required, performance. The, from the, Everpure standpoint, each FlashArray controller, as usual, must be connected to the, network switches, and internally there will be a couple of, virtual switches that Nutanix, in turn, allocates for, the, AHV environment,
specifically for the NVMe/TCP, traffic. From the archit- from the P- Everpure architecture standpoint, all the, that, all that is involved are realms, ports, the cost and volumes, plus the, as usual, the snapshots and protection groups. Minimum requirements are, from Purity six to 10 to three.
What, what is required here is the, and which is, also available, which is made available by the latest six dot 12, six dot 12 release, is the, tight integration with the, the clu- the, ch-change block tracking mechanism that was not available before on, on the Prism. On, AHV, seven dot five dot one, which is the first release that was,
available for the integration, and in combination with the six dot 10, which was not having that feature as well. So, with the, the latest version, seven dot six, plus, Nu- Nutanix seven dot six plus Purity six, 12 plus, there is this tight integration between, the, in- internal, ch- CBT, mechanism and, the,
Nutanix volume mapping. Minimum, the, the caveat, for the moment is that, for, each Nutanix cluster there, has to be a single, one single port on the FlashArray. There is anyhow the possibility to have many ports per realm. Nutanix requires a specific realm on the FlashArray for that, with that single port.
But a single array can serve m- multiple clusters. Then there is, one specific user that has to be dedicated to the realm for, the external storage connection to be configured in the array. For, e- everything pivots, around the Nutanix Stargate for the, translation between, NVMe/TCP, towards, the array, array side and
internally to, the, each single virtual machine. So Stargate as the gateway, fr- from iSCSI down to, NVMe/TCP to each single, external disk. The limitation, for the Nutanix cluster supports the ten thousand virtual disks, which corresponds, of course, one, one, one by one, to, ten thousand,
Purity, FlashArray volumes. Four thousand FlashArray per host are the limitation for, the, the Purity. And as I said, the connectivity, pivots around the Stargate. The, each single Stargate in a cluster, is rotated, in, in the fashion that you, you see listed there. So allocation starts from Stargate one, then
once Stargate one fills up, with its connections, the, the Stargate do- two is involved and so on. Minimum NCP, requirements, as I said, Prism Central si- si- seven dot five, so everything again, stays around, seven dot five plus, which was the first release, that, accommodates for external storage and has the integration with the, Pure
FlashArray. A minimum of three nodes and of course, if you upgrade it to seven dot six, you will benefit of whatever wa, wa- what I was saying before. The direct mapping between the internal, cloudbo- block tracking mechanism in, change block tracking mechanism in Pure and the, Prism APIs. From Everpure side, six dot, 10 is the minimum
requirement, is, FlashArray, X, XL, C and, RC are all supported. What is crucial, is the NVMe/TCP, service with the two, plus, five, tw- twenty-five Gigabit Ethernet cards. So the minimum requirement is just with two, twenty-five Gigabit Ethernet cards, and that goes, as I said at the beginning, the main goal for, having external storage connected to
a Nutanix array is to accommodate for high IO, IO demanding workloads. Let's go straight to the, interesting part, the FlashArray configuration, and here, the, configuration can happen through the, CLI or through the GUI. If you have a, a six, dot twelve that's happened too easily through the GUI. You have to configure the, NVMe TCP card first with the usual pure
network command. Then you have to create a realm for, specifically for, the Nutanix cluster you want to connect. Once you have your, your realm created, you add the pod, so single pod for, the array, for the cluster. You can optionally set a quota limit, the-- if you wa-- if, if you want.
And then a policy, this is, this, is, what the policy is required for the realm so that the Prisma can communicate directly with the FlashArray using the credential assigned to the specific user. So, for the sake of the demo, the, the there is a pre-recorded, the, pre-recorded video that I'd taken for the configuration, and that goes through the
preparation of the array, the, storage with switches on, the, Nutanix side, the connection in Prisma. Then, I'll, I'll take you on a live demo for pro-- just to showcase how a, a VM can be provisioned and, how to how, snapshots can be directly taken from, the Prisma, UI and what they generate on the array side.
Those of-- all those of you that are used with the, Everpure arrays will notice that, the, immediately that the volumes, created on the, on the array are corresponding one-to-one to the vi- virtual machine disks. Same is for the snapshots. The pure snapshots, are the Nutanix, managed snapshot originate as a,
a pure snapshot on the FlashArray. So this is the creation, of the, realm on the FlashArray. The, demo shows how to create things directly from the console. As I said, this is, feasible as well from, the, UI, from the Pure UI, UI. So first thing, create the realm, then the pod,
which is dedicated to Nutanix. Then we have the pod. Then now the policy creation for the specific realm and user. And once the user is created, that, that is all is needed from the array side. You may want to verify that everything works from the, user that has been created so that
you don't have to experience any issue when you connect from the Prisma UI. So the pod is there. We can set the quota limit, as we said before. And here on, the, prism element, you can
configure the-- you have to configure the virtual switches. You create, two virtual switches, one for each physical switch. It's highly recommended to use, jumbo frames here. The, You add the virtual interfa-- the physical interfaces to the
virtual switch. Paying attention, those are the twenty-five gigs nic, n-nicks that are required. Then you create a second switch in addition to that. You have to f- spec-specifically tag that for the external storage. So, and then you have to repeat the same operation for the second, for the second
virtual switch Assign the jumbo frame as well, and here you are with the, the configuration that is ready for, attaching external storage. So this is the part for ne-necessary for the network setup. Now you have to move on, set up, setting up, external storage.
And here you can select one of those, that are available for Nutanix. Give you a name, the IP address and the username you created before for, the, the realm. Then you can select here the realm and assign-- the, assigned pod. Attach that array and everything from the provisioning standpoint is taken over by
the, Prisma UI. So you get the array. As you can see, this was a, system, brand-new system, brand-new, AHV cluster without any configured storage. The external storage, m- it's mandatory to connect external storage to,
Nutanix, environments that are not provisioned with internal storage. So you cannot have a, a, a Nutanix storage coexisting with the, external storage. And here you see that, the provisioning mechanism has, created if, created the, the volumes on the, array side. From, the volume standpoint in the arrays, in the array, what, what, what you end up is
with the, data volume for each single VM. You can find sometimes that there is also a meta disk volume. The case, disk is the case whenever you operate through, a, a Nutanix replication, which, creates a metadata volume f-for each single VM, a metadata volume for each si-single VM.
But normally, what you have for a VM is just the DT volume, and in, the, in case of a, a cloned VM, you have the C, VDisk, data volume. The metadata volume is something that, you only will find in situations where Nutanix, disaster recovery, disaster re-- Nutanix replication is in place.
Then there is, the, the volume for There are the volumes for the Zookeeper, for the Curator and Cassandra. We will go, we will see this immediately on the array. And, if we look at the, at the actual situation on the array, we go here and we have for the Nutanix pod, in this case, this is my Nutanix
environment realm with its pods. Here, as you can see, there is only, DT volumes. This is the result of the migra-- the, upgrade to seven dot six. There is not any, MD volume here, so the-- and they are corresponding, to, each of the single components.
Most of them are related to the virtual machines that are running there, but there is also volumes that are, as you, as we, we can see here, this is a volume for the Curator. The-- and there are other volumes for Zookeepers, for Cassandra. All those are mainly provisioned by, the installation procedure.
And yeah, here, for instance, we have volumes that are for the Zookeeper, instances. The, all the other da-- data volumes are mainly for the virtual machines. So if we look at things, at this point on this environment where we have Prisma Central installed
and also Prism. What we can do in this environment very easily and quickly, we can create a virtual machine. For the sake of the demo, I already created one here, but create a new one Let's assign four CPUs, eight gigs of RAM. Add a new disk. Here I'm going to select
one image. I can use Rocket 10. Add. Then I can add my network card here. Add. Save. Okay, here we are with our machine.
What we should see here, 'cause the easiest way for finding a mapping from the Prisma element that you don't find easy, the, to fi- to, the, association between the virtual disks and, the, disks on the array. But here if you go on the VMs on Prisma Central, that's pretty easy to do. We can click on the VM, look at the disks, and here is our disk.
We can, This is not visible in the Prisma UI, but if we go here, the mapping is pretty straightforward. Same thing is for the APIs. So this is our 10 gig disk that has been taken, that's, has been created automatically by the, Prisma, element UI, when, the, well, actually through the APIs when
I created the VM. Here just for, since, I'm, I'm close to the end, what I can do, I can, take a snapshot just to sh- show. I can also start the VM meanwhile. Power on meanwhile, so.
And here is the snapshot that I've just taken. So this shows, how, direct and easy it is for, the Nutanix environment to integrate with the a Pure FlashArray. We should see in a while that the machine is booting up. Yeah, and here we are.
Yeah, I can take another snapshot right now from this just to And we should see that there is a second snapshot here. Yep, yep, and we have the two snapshots. Well, that, that's it for this demo, 'cause I'm close to the half an hour, and, the de- de- the demo was, expected to, just span across, half
an hour. Couple of considerations more just to close the session as I promised. Maximum number of VMs for the array is 50. Those are normal limits for the array. And if you look at the, Nutanix side, so back to this us- anything different than usual,
so maximum number of host. This is wrong. It was 4,000. It's not 2,000. The, no, no. And, and here for the Nutanix standpoint, which is probably the most important, the
minimum number of nodes for a production environment is three. So this is, good for existing, Nutanix environment that span across multiple nodes. So it is possible to carve a minimum number of nodes for an existing environment and dedicate those to, high-performing workloads that have to be accommodated on
a, Pure, FlashArray. Maximum number of virtual disk per cluster is 10,000. The, the number of, of VMs that are using a sync replication per cluster, this is the Nutanix synchronous replication, is 500. Compatibility, as I said, everything starts from 7.5 for the, Cloud Manager, for the live
migration, all these things that you see here, any, Nutanix, Nutanix, cloud platform. Kubernetes platform is supported. Things that are not supported yet are NDB, NUS, and NDK. NDB is, is, is ongoing, the, supportability for that. The NVMe-is, are things that, you may want to, evaluate case by
case with your Nutanix representative for this kind of integration. And for the support standpoint, Nutanix is responsible for their side of, the, the product. But whatever hap- happens, is jointly routed to Pure//E if it is discovered that the problem re- resides on the, on the storage side.
P- Everpure is, responsible as usual for the, Purity and hardware support, and we work tightly, with the Nutanix for everything that, pertains to the support cases. Next steps, from the Pure, from those that are Pure customers, the easiest way is to, recur to our Test-Drive. There is a Nutanix, instance that is available, on Test-Drive that
you can spin up. You can, the, there is, the documentation that drives you to completely configure it from scratch, like, same steps I, I shown. Then, the POC as usual. So that's it for this kind of presentation. Thanks for, attending, and, see you to the next, Tech Talk.
Eugenio Grosso
Principal Field Solutions Architect, Everpure
Traditional hyperconverged infrastructure ties compute and storage together, forcing you to add nodes you don't need just to gain capacity. This webinar shows infrastructure and storage admins how pairing Everpure FlashArray with the Nutanix Cloud Platform over NVMe/TCP breaks that dependency, delivering external, high-performance block storage to Acropolis Hypervisor (AHV) clusters.
You'll walk through the reference architecture in detail, including how the two platforms connect and how data services like VM-level snapshots, replication, and SafeMode ransomware protection layer beneath your Nutanix workloads.
The result is an architecture you can scale on your terms, expanding compute and storage independently while maintaining the enterprise data services and non-disruptive operations your workloads depend on.
Key takeaways:
- How FlashArray connects to Nutanix AHV over NVMe/TCP, and what the end-to-end reference architecture looks like in practice
- Why disaggregating compute and storage eliminates node overprovisioning and lowers TCO without sacrificing performance
- How to extend enterprise data services — snapshots, replication, and ransomware-resistant immutable snapshots — beneath your Nutanix workloads
- What non-disruptive upgrades and independent scaling mean for day-2 operations and long-term capacity planning
We Also Recommend...
Personalize for Me
My Updates
No updates at this time.