A customer asked me recently a deceptively simple question: “If we put our most important applications on Nutanix, what actually matters?” Not the datasheet, not the feature matrix, but the handful of capabilities that make the difference when it’s your ERP, your core banking database or your electronic health record system that has to stay up, perform and be recoverable. I’ve spent a big part of my career virtualising business-critical applications, first on VMware and for more than a decade now on Nutanix, so this is my practical list of the ten capabilities of the Nutanix Cloud Platform that I think matter most, and why.
1. Performance and Data Locality
Business-critical apps are rarely bottlenecked on CPU. They’re bottlenecked on storage latency, and more importantly on latency consistency. The AOS distributed storage layer is built around data locality: when a VM writes data, one copy is stored on the node where the VM is running, so reads are served locally rather than across the network. When a VM migrates, its data follows it over time. For an Oracle or SQL Server database doing lots of small random reads, that’s a big part of why latency stays low and predictable.
On all-NVMe nodes, AOS uses Blockstore with SPDK, which moves device I/O into user space and removes kernel storage driver overhead from the I/O path. Add more nodes and you add more controllers, and more drives, so performance and capacity scale out together rather than hitting a dual-controller ceiling. I’ve written about this in practice before, for example Nutanix performance with Oracle SLOB on all-flash nodes. As always, size for the workload and validate with your own tests, but this architecture is what makes the platform a good fit for Oracle, SQL Server, SAP HANA and EHR systems such as Epic.
2. Resilience Built In, Not Bolted On
Every write in AOS is protected by a replication factor: RF2 keeps two copies, RF3 keeps three (RF3 needs at least five nodes). Erasure coding (EC-X) can then reduce the capacity overhead for colder data without giving up protection. Availability domains add node, block and rack awareness, so replicas are placed to survive the loss of a whole chassis or rack, not just a disk. The data resilience chapter of the Nutanix Bible covers this in detail.
What I like most is how rebuilds work. When a drive or node fails, every remaining node takes part in re-protecting the data, so the more nodes you have, the faster you’re back to full redundancy. There’s no hot spare and no single RAID group grinding through a rebuild for hours. On top of that, Life Cycle Manager (LCM) handles software and firmware upgrades with automated dependency checks, and AOS upgrades roll through one Controller VM at a time while I/O is redirected to the other nodes. For apps where you can’t get a maintenance window, that matters.
3. Data Protection and Disaster Recovery
Native snapshots are VM-centric and can be application consistent using Nutanix Guest Tools and VSS, so you can quiesce the database rather than hoping a crash-consistent copy is good enough. From there you pick the replication tier that matches each application’s RPO, as described on the Nutanix Disaster Recovery page:
- Asynchronous replication: RPOs from 1 hour to 24 hours, suitable for most workloads and long distances.
- NearSync: RPOs of 1 to 15 minutes, using lightweight snapshots, without the latency and distance limits of synchronous replication.
- Synchronous replication and Metro Availability: zero RPO between sites within metro distances. On AHV this is AHV Metro with a witness for automated failover. On ESXi it’s Metro Availability, which builds on stretched clustering.
Nutanix Disaster Recovery (which many of you will remember as Leap) is managed from Prism Central. Protection policies are applied by category, and recovery plans act as runbooks that define boot order, network mapping and scripts. You can also run non-disruptive test failovers, which is the bit most DR plans skip. The target doesn’t have to be another data centre. You can fail over to Nutanix Cloud Clusters (NC2) in the public cloud, and Multi-cloud Snapshot Technology can send snapshots to S3-compatible object storage. The data protection and DR solutions page also lists the backup partners that integrate with these native capabilities, such as Veeam, HYCU and Commvault.
4. Certified and Validated for Enterprise Applications
Certification and vendor support are usually the first questions from application owners, and rightly so. Nutanix is certified for SAP HANA and SAP NetWeaver, and was the first hyperconverged platform certified for both. I covered some of that history in SAP HANA validated for production on Nutanix AHV. Oracle Single Instance and RAC, Microsoft SQL Server (including Always On Availability Groups and Failover Cluster Instances), and healthcare EHR platforms such as Epic, MEDITECH and Oracle Health (formerly Cerner) all run on Nutanix in production. The healthcare solutions page covers the EHR side. As always, check the specific vendor’s support statement and the Nutanix best-practice guide for your exact version and configuration.
The bigger operational win is Nutanix Database Service (NDB), which automates provisioning, patching, cloning, copy data management and database protection for SQL Server, Oracle, PostgreSQL, MySQL and MongoDB. For Oracle it can deploy Single Instance, RAC and Data Guard configurations. For SQL Server it can deploy standalone instances, Availability Groups and FCIs. Thin clones are a big time saver for refreshing non-production environments. I showed a real example in refreshing a 100TB production Oracle DB to ten non-prod DBs in 20 minutes.
5. Choice of Hypervisor and Hardware
AHV isn’t sold as a separate product. It’s included with the Nutanix Cloud Infrastructure (NCI) licence, so there’s no separate hypervisor licence to buy. VMware ESXi is also supported on the same storage layer, which lets you move at your own pace. On the hardware side, you can run on Nutanix NX or on OEM and partner platforms from Dell, HPE, Lenovo, Cisco, Fujitsu and others. The hardware platforms page links to the compatibility lists.
AHV is also an NVIDIA-Certified Hypervisor. In October 2026 Nutanix announced that AHV had been certified under the NVIDIA-Certified Hypervisors program, which validates that a virtualisation platform delivers near-bare-metal performance for representative AI and accelerated computing workloads. NVIDIA’s certified hypervisors list shows Nutanix Cloud Infrastructure 7.6.9 with AHV 11.3 and all later AHV 11 releases, certified on the Hopper NVL8 platform. AHV supports both GPU passthrough and vGPU, and AHV is listed in NVIDIA’s vGPU software support matrix, so the same hypervisor that runs your databases can also run the GPU-accelerated AI workloads as described in section 10.
For many customers right now, this is the practical path off a hypervisor whose licensing has changed underneath them. Nutanix Move is a free migration tool that moves VMs from ESXi, Hyper-V and various cloud platforms onto AHV. It does incremental data seeding and then a short, scheduled cut-over, and no application refactoring is needed. For business-critical systems, plan the cut-over like any other change, with a tested rollback path.
6. Hybrid Multi-cloud Portability
NC2 runs the same Nutanix stack on bare-metal instances in AWS, Azure, Google Cloud and OVHcloud. It’s the same AOS, the same AHV and the same Prism, so the operating model, the runbooks and the skills carry across, and the apps don’t need refactoring. Licences are portable between on-premises and the public cloud, with pay-as-you-go options as well. That makes NC2 useful for DR targets, data centre exits, bursting for seasonal peaks, or getting apps closer to cloud-native services without re-platforming them first. The hybrid multicloud page covers the broader strategy. I wrote about an early version of this pattern in Hybrid Multi-Cloud Onramp with Nutanix Cloud Platform on Equinix Metal.
7. Unified Storage: Files, Objects and Volumes
Business-critical apps rarely live on block storage alone. There are file shares for application data and interfaces, object storage for backups and archives, and sometimes bare-metal or guest-clustered workloads that need iSCSI. Nutanix Unified Storage (NUS) provides all three on the same platform:
- Files: SMB and NFS file services.
- Objects: an S3-compatible object store with WORM policies and versioning, so backup copies can be made immutable. That’s an important last line of defence against ransomware.
- Volumes: iSCSI block storage for VMs, physical hosts and guest clustering. I covered this in simple guest OS clustering without complex config.
NUS is also built for AI data pipelines. In the MLCommons MLPerf Storage v2.0 benchmark, a single 16-node Nutanix Unified Storage cluster serving data over NFS kept 2,312 simulated NVIDIA A100 accelerators fed in the ResNet-50 training workload (see the published results and Nutanix’s write-up). NUS is also NVIDIA-Certified Storage at the enterprise level and is listed on NVIDIA’s certified storage page. It uses parallel NFS and GPUDirect Storage over NFS with RDMA to move data directly between storage and GPU memory, so the same platform can feed the AI workloads covered in section 10.
Using one platform means one set of skills, one support call and one lifecycle to manage, rather than three storage silos.
8. Security and Compliance
Security for these apps has to be on by default and provable to an auditor. The Nutanix security stack starts with a hardened platform. The Controller VMs and AHV are built to a security baseline aligned with DISA STIGs, and Security Configuration Management Automation (SCMA) regularly checks the configuration and automatically corrects drift. Data-at-rest encryption is available in software (AES-256) or with self-encrypting drives. You can use the built-in native key manager, so a third-party KMS isn’t mandatory, or integrate with your existing enterprise key management.
For the network, Flow Network Security provides application-centric micro-segmentation with default-deny policies that limit lateral movement. That’s exactly what you want around a database tier. For unstructured data, Data Lens analyses access patterns to detect and block ransomware activity on Nutanix Files shares. On the compliance side, the compliance and certifications page lists FIPS 140 validated cryptographic modules, Common Criteria certification and DoDIN APL listings for AOS, AHV and Files. Note that encryption and micro-segmentation depend on your licence tier or add-ons, so check what you’re entitled to.
9. Operations and Automation
The day-two experience is where platforms either save you time or cost you time. Prism Central provides a single place to manage many clusters across sites and clouds. NCM Intelligent Operations (formerly Prism Pro/Ultimate) adds capacity runway forecasting, what-if planning, anomaly detection, rightsizing of over- and under-provisioned VMs, and low-code Playbooks for automated remediation. Capacity planning is something every business-critical app owner should care about, because running out of headroom is an outage waiting to happen.
NCM Self-Service (formerly Calm) lets you publish blueprints for complete multi-VM applications through a marketplace or ServiceNow, with governance and quotas built in. Underneath it all are REST APIs and SDKs documented on nutanix.dev, along with an official Terraform provider and Ansible collection. So your database and app platforms can be built as code and rebuilt the same way every time. Predictability and consistency is a key aspect of operating business-critical applications.
10. Modern Apps and AI Alongside Your VMs
Business-critical apps are increasingly a mix of traditional VMs and containerised services, and more and more often they include an AI component that needs access to the data in those systems. Nutanix Kubernetes Platform (NKP) is a CNCF-conformant Kubernetes platform with fleet management and GitOps that runs on Nutanix, other infrastructure and the public cloud. Nutanix Data Services for Kubernetes (NDK) adds application-centric snapshots and replication for stateful containers, and NUS provides persistent storage, so containers and VMs share the same platform and the same protection model.
For AI, Nutanix Enterprise AI (NAI) lets you deploy and govern LLMs from NVIDIA NIM, Hugging Face or your own models behind secure, managed inference endpoints. It runs on NKP or on EKS, AKS and GKE. NAI is the core of the validated GPT-in-a-Box 2.0 stack described in the NAI FAQ. Running inference next to your business-critical data, rather than copying that data out, has big security and data-gravity advantages. I went deeper on that in Bring the AI to the Data – Business Critical Databases Meet Enterprise AI on Nutanix. Because AHV is an NVIDIA-Certified Hypervisor (see section 5), GPU-accelerated inference can run in VMs on the same platform as the data.
Final Word
None of these capabilities is unique on its own. What makes the difference for business-critical applications is having all of them in one platform with one operating model: consistent low latency from data locality, self-healing resilience, a choice of RPO from 24 hours down to zero, the certifications that application owners ask for, and the freedom to choose your hypervisor, your hardware and your cloud. That removes a lot of the integration risk that used to come with building infrastructure for tier-1 apps. If you’re evaluating a move, start with the application’s real requirements for RPO, RTO, latency, licensing and support, and map them to the sections above. Then prove it with a proof of concept on your own workload. Some features depend on your licence tier, so check the software options and licensing page as part of that work.
If you’d like more detail on any of these areas, let me know in the comments or on X. I’m always happy to talk business-critical apps.
This post first appeared on the Long White Virtual Clouds blog at longwhiteclouds.com. By Michael Webster. Copyright © 2012 – 2026 – IT Solutions 2000 Ltd and Michael Webster. All rights reserved. Not to be reproduced for commercial purposes without written permission.