Back in July, I wrote up the PowerStore GA on Nutanix AHV: what shipped, what you need, and how to get started. That post didn’t address whether you should deploy it this way at all.
That is the harder question, and it is not one a launch announcement is built to answer. So here is the version I give customers: where PowerStore-backed Nutanix reaches its limits, where classic HCI is still the better buy, and how the licensing actually works.
What are you actually deploying?
The CVM does not go away, and this is the part most people get wrong. AOS still runs, Prism still manages, and Flow still handles microsegmentation, etc.
What changes is where the data is. Instead of the Acropolis Distributed Storage Fabric (ADFS) pooling local HDD/SSD/NVMe across your nodes, every VM vdisk maps 1:1 to a volume on the PowerStore. Stargate talks to the array over NVMe/TCP through a user-space SPDK initiator, and a cluster service called Castor handles the array’s lifecycle. Even the cluster’s own metadata services, meaning Cassandra, ZooKeeper, and Curator, live in service volumes on the array rather than on local disk.
Two consequences follow, and everything below is downstream of them. Your storage services, meaning snapshots, clones, data reduction, and replication, are PowerStore’s rather than ADSF’s, with Prism orchestrating and the array executing. And your storage network becomes the storage bus.
Where does the integration reach its limits?
One Appliance Per Cluster. As of AOS 7.6, PowerStore must be a single-appliance cluster. PowerStore itself scales out to four appliances, but Nutanix does not consume that yet. If you need more than that, you are adding a second cluster and a second array rather than growing the one you have.
NVMe/TCP or nothing. There is no Fiber Channel support, and Nutanix has been clear that it has no plans to add it in either the near or long term. If your storage environment is Fiber Channel today, this integration will not reuse your existing fabric.
Failure Domain. With ADSF, a node loss affects only a fraction of the cluster, and the rebuild runs in parallel across the remaining nodes. PowerStore places two controllers in a 3U chassis, so losing one means the survivor handles the whole cluster’s I/O until service is restored.
Nutanix data services you do not get on that cluster. Files and Objects run on ADSF. On an external-storage cluster, they aren’t available, so if you need SMB shares or S3 alongside those VMs, you need a separate NUS deployment on separate hardware. The same applies to the ADSF features that appear in many TCO models, including erasure coding, ADSF deduplication, and compression. PowerStore performs its own data reduction, backed by Dell’s 6:1 guarantee on Gen 3 arrays running PowerStoreOS 5.0 or later, which is perfectly good and arguably better.
Check the support matrix. External storage support has been expanding release by release. Capabilities unsupported six months ago are supported now, and a few combinations still aren’t. Verify EUC/VDI and any specialty workload against the current support matrix rather than against a blog post, including this one.
Two lifecycles, two support models. Nutanix LCM owns AOS and AHV, while Dell owns PowerStoreOS. Dell Private Cloud and the Dell Automation Platform bring those together with validated blueprints and system-level support, which is the real reason to buy it this way rather than assemble it yourself. You are still tracking an interoperability matrix in two directions. Every AOS upgrade becomes a question about PowerStoreOS, and every PowerStoreOS upgrade becomes a question about AOS.
The entry requirements are higher than people expect. Minimum of three compute nodes, PowerStore Gen 1 through Gen 3 (excluding PowerStore X), PowerStoreOS 5.0.0.2 or later, AOS 7.6, AHV 11.2, Prism Central 7.6, Foundation Central 2.2, and Move 6.3.0 for migrations. An organization sitting on AOS 6.x with a Gen 1 array is not a short step away from this.
When to go PowerStore?
There are shapes of problems where disaggregation is clearly the better answer, and I have walked into several of them:
- Lopsided growth. Capacity climbing while cores sit idle, or the reverse. The common version is buying nodes for terabytes and paying for cores you will never schedule a VM on. Disaggregation removes that tax outright, and that alone can pay for the additional switches.
- An array with runway left. Moving off your current hypervisor without also writing off the array is a legitimate reason to do this, even when HCI would be the cleaner greenfield design.
- All-inclusive storage licensing. PowerStore’s software is bundled, so there are no per-feature storage SKUs to model or overlook.
When to go HCI?
In most cases, and specifically in these:
- Small sites and edge. Three nodes, two switches, and you are done. Adding an array and two more switches to a three-node site roughly doubles the parts count and the failure modes while buying you nothing.
- Linear, predictable growth. If compute and capacity grow together.
- You want Files, Objects, or a genuinely mixed-services cluster. Unified Storage on ADSF remains one of the strongest reasons to stay converged.
- You are growing past a single appliance. This returns to the ceiling above. ADSF simply keeps taking nodes.
- You want one number to call. Dell Private Cloud narrows this considerably, but converged remains fewer moving parts and fewer conversations.
This is roughly how the conversation tends to go on a whiteboard:
None of these is a single disqualifier. The point is that the answers tend to cluster. When I have seen them split evenly down the middle, that is usually a sign the cluster is trying to do two jobs and should be two clusters.
Licensing
On the Nutanix side, there are two ways to license compute-only nodes that sit on external storage, and the choice matters more than it might at first appear.
NCI is the license you probably already own. It is core-based, portable, has no core-count minimums, and can move between HCI clusters, external storage clusters, and NC2 in the cloud. If your environment is mixed, or you think it might be within three years, that flexibility is usually worth more than the difference in list price.
NCI-C (NCI Compute) is priced for compute-only nodes. It is the right answer when the environment is built entirely on external storage, and there is no realistic scenario in which those licenses need to return to an HCI cluster.
The mistake I see most often is that teams model NCI-C economics against a roadmap that quietly assumes HCI somewhere, whether a Files cluster, an edge site, or a development environment. Choose the license that matches where the workloads actually live in year three, not year one.
Summary
PowerStore on Nutanix is a good answer to a specific question: how to decouple compute from capacity, or keep an array you have already paid for, without giving up AHV and Prism. If that is your question, this is one of the cleanest options on the market at present, and Dell Private Cloud makes it a supported product rather than an assembly exercise.
If it is not your question, proceed carefully. The single-appliance ceiling, the Ethernet-only storage fabric, the loss of Files and Objects on that cluster, and the two-lifecycle operating model are all real costs, paid by the people who run the platform rather than the people who sign for it. HCI remains the default for small sites, for linear growth, and for anyone who would rather not run a storage network. Disaggregation is an exception you deliberately reach for, not an upgrade.
Sources:
- Nutanix: The Nutanix Bible, Book of External Storage: Architecture; Tech Center, Migrating to NVMe/TCP; Community, Unlocking Synchronous Replication with AOS 7.5.1; Licensing and Cloud Platform Software Options.
- Dell: KB 000226344, PowerStore Supported and Unsupported Cluster Configurations; Dell Private Cloud: Deploying Nutanix (technical brief, PDF).
- Independent: DCIG, Nutanix’s Support For External Storage Systems Comes With Qualifications (April 2026); StorageReview, Dell PowerStore Gen 3.
Related reading: PowerStore on Nutanix AHV Is Here covers what shipped and the full requirements list. Nutanix + Dell PowerStore: Expanding Storage Choice for the Hybrid Cloud is the original 2025 announcement write-up.
Have you deployed this yet, or decided against it? I would like to hear which way it went and why.
Let me know below.
Leave a Reply