Nutanix Cloud Platform Overview¶

Most enterprises still run their workloads on a tangle of separate systems — one vendor for compute, another for storage, another for networking, yet another for virtualization. Managing all of that is expensive, slow, and fragile. Nutanix was founded on one radical idea: collapse all of those layers into a single, software-defined platform that runs on commodity hardware and is as simple to operate as a public cloud.

In 2026, Nutanix Cloud Platform (NCP) has grown from that original idea into a comprehensive stack spanning private cloud infrastructure, multi-cloud management, enterprise Kubernetes, database-as-a-service, AI infrastructure, and unified storage — all managed through a single pane of glass.

Part 1: The Problem Nutanix Was Built to Solve¶

The Traditional Datacenter Tax¶

Before Nutanix, a typical enterprise datacenter looked like this:

┌─────────────────────────────────────────────────────────────┐
│  Compute Tier     │  Blade servers, rack servers            │
│  (VMware/Hyper-V) │  Separate vendor, separate support      │
├─────────────────────────────────────────────────────────────┤
│  SAN/NAS Storage  │  Dedicated storage arrays               │
│  (EMC, NetApp)    │  SAN fabric, HBAs, FC switches          │
├─────────────────────────────────────────────────────────────┤
│  Network Fabric   │  Top-of-rack switches, load balancers   │
│  (Cisco, Juniper) │  Another vendor, another team           │
└─────────────────────────────────────────────────────────────┘

Every layer had its own management interface, its own support contract, and its own failure domain. A storage performance problem required a storage engineer, a compute engineer, and a network engineer in a room together — just to diagnose it.

Public clouds like AWS showed a better way: compute, storage, and networking delivered as one integrated service with self-service provisioning and pay-as-you-go scaling. Nutanix brought that operating model on-premises.

The Hyperconvergence Answer¶

Hyperconverged Infrastructure (HCI) means collapsing compute, storage, and virtualization into a single software layer that runs on standard x86 servers. Instead of a separate storage array connected via SAN, each server node contributes its local disks to a shared, distributed storage pool managed entirely in software.

The result: - No external storage arrays - No SAN fabric or HBA cards - No separate storage management team - Scale by adding nodes — compute and storage grow together

Part 2: The Architecture Foundation — AOS¶

What AOS Is¶

AOS (Acropolis Operating System) is the distributed storage and platform OS that powers every Nutanix cluster. It runs as a Controller VM (CVM) on every node — a dedicated virtual machine that handles all storage I/O for that node and coordinates with CVMs on other nodes.

┌────────────────────────────────────────────────────────────────────┐
│                        Nutanix Cluster                             │
│                                                                    │
│  ┌──────────────┐   ┌──────────────┐   ┌──────────────┐           │
│  │    Node 1    │   │    Node 2    │   │    Node 3    │           │
│  │              │   │              │   │              │           │
│  │  VM  │  VM  │   │  VM  │  VM  │   │  VM  │  VM  │           │
│  │──────────────│   │──────────────│   │──────────────│           │
│  │     CVM      │   │     CVM      │   │     CVM      │           │
│  │  (AOS core)  │   │  (AOS core)  │   │  (AOS core)  │           │
│  │──────────────│   │──────────────│   │──────────────│           │
│  │ SSD │  HDD  │   │ SSD │  HDD  │   │ SSD │  HDD  │           │
│  └──────────────┘   └──────────────┘   └──────────────┘           │
│                                                                    │
│            ←──── 10/25/100 GbE Internal Network ────→              │
│            ←──── Distributed Storage Fabric (DSF) ────→            │
└────────────────────────────────────────────────────────────────────┘

The CVMs communicate over the standard network (no special SAN hardware required) and together form the Distributed Storage Fabric (DSF) — a single logical storage pool that any VM in the cluster can use.

The Controller VM Design¶

The CVM is one of Nutanix's most important architectural decisions. Instead of building storage logic into a kernel driver (which would be tightly coupled to the hypervisor), Nutanix put the entire storage stack into a user-space VM. This gives three major advantages:

1. Hypervisor independence: The same CVM runs on AHV, ESXi, or Hyper-V. You can even switch hypervisors without changing your storage.
2. Upgradability: The storage software can be upgraded independently, live, one node at a time — no maintenance windows.
3. Isolation: A storage bug cannot crash the hypervisor. The VM is a fault boundary.

Data Resiliency: Replication Factor¶

Every write in AOS is replicated across multiple nodes for resilience. Nutanix calls this the Replication Factor (RF):

For metadata (the index that tracks where every data block lives), Nutanix uses a separate Fault Tolerance (FT) setting:

• FT1: 3 metadata copies — minimum 3 nodes required
• FT2: 5 metadata copies — minimum 5 nodes required

Data Locality: The Performance Secret¶

Here's the detail that makes Nutanix fast by default: one copy of every data block is always written to the local node where the VM is running.

VM on Node 1 writes data
         │
         ▼
┌─────────────────┐     ┌─────────────────┐
│     Node 1      │────▶│     Node 2      │
│  LOCAL COPY ✓   │     │  REMOTE COPY    │
│  (SSD, fast)    │     │  (for RF2)      │
└─────────────────┘     └─────────────────┘

Read path: served from local SSD — no network hop needed

Reads are served from the local copy at SSD speed without crossing the network. This is why Nutanix performs well even on 1 GbE networks that would cripple a traditional SAN-based setup.

Part 3: The Hypervisor — AHV¶

What AHV Is and Why It Exists¶

AHV (Acropolis Hypervisor) is Nutanix's built-in, enterprise-grade hypervisor. It is based on Linux KVM (Kernel-based Virtual Machine) — the same battle-tested hypervisor technology used by major public clouds — enhanced with:

• QEMU 6.2 for hardware emulation
• Libvirt 8.0 for VM lifecycle management
• Open vSwitch 2.14 for distributed virtual networking

As of 2026, 69% of Nutanix clusters run AHV — up from a small minority just a few years ago. The growth is driven by cost: AHV is included with Nutanix at no additional license fee. For customers running VMware vSphere, switching to AHV eliminates a significant hypervisor cost.

What AHV Provides¶

┌───────────────────────────────────────────────────────────┐
│                     AHV Capabilities                      │
├────────────────────┬──────────────────────────────────────┤
│ VM Management      │  Create, clone, snapshot, migrate     │
│ Live Migration     │  vMotion equivalent — zero downtime   │
│ HA & DRS           │  Automatic restart + workload balance  │
│ Virtual Networking │  VLANs, bonds, Open vSwitch fabric    │
│ GPU Support        │  NVIDIA vGPU passthrough for AI/VDI   │
│ Security           │  Secure boot, vTPM 2.0 support        │
│ Migration Tools    │  Move from VMware/Hyper-V to AHV      │
└────────────────────┴──────────────────────────────────────┘

AHV is managed entirely through Prism (Nutanix's management plane) — no vCenter equivalent required. Fewer management layers mean fewer failure points and lower operational complexity.

Part 4: Storage Efficiency¶

Raw storage capacity costs money. Nutanix includes several technologies to squeeze more usable capacity out of each node without sacrificing performance.

Tiered Storage (SSD + HDD)¶

Each node typically has SSDs for hot data (frequently accessed) and HDDs for cold data (rarely accessed). AOS continuously monitors access patterns and automatically migrates data between tiers — hot data moves to SSD, cold data moves to HDD. This happens transparently, with no administrator involvement.

Erasure Coding¶

Standard RF2 keeps 2 full copies of every byte — a 2x storage overhead. Erasure Coding (EC) is a post-process technique that replaces full copies with mathematical parity stripes, similar to RAID-6 but distributed across nodes.

EC is applied post-process (after writes) so it never adds latency to the write path. The result: you can store roughly 60% more data in the same hardware compared to RF2 without EC.

Deduplication¶

The Elastic Deduplication Engine (EDE) fingerprints data at 16K granularity. When two blocks are identical (common in VDI environments with many VMs cloned from the same base image), only one copy is stored and a pointer is kept for the duplicate. This works across the entire cluster — not just within a single node.

Compression¶

AOS supports both: - Inline compression (LZ4 algorithm) — compresses data before writing to disk, best for incompressible data environments - Post-process compression — applied during idle periods to minimize impact on production I/O

Snapshots and Clones¶

Nutanix uses redirect-on-write for snapshots, which means: - Snapshots are instant (no data is copied at snapshot time) - Reading from a snapshot reads original blocks directly - Writes after a snapshot go to new blocks; old blocks remain unchanged - Snapshots consume space only for changed data

Clones use the same mechanism — a full VM clone is created in seconds and initially shares all storage blocks with the original. It diverges only as changes are made.

Part 5: Disaster Recovery¶

Data protection in Nutanix spans four distinct RPO/RTO tiers, all manageable from Prism Central with 1-click failover.

RPO Timeline
──────────────────────────────────────────────────────────
0 sec ──── 1 min ──── 15 min ──── 1 hr ──── multi-hr
  │              │                  │
  │ Synchronous  │ Near-Sync        │ Async Replication
  │ (Metro)      │ (RPO: ~1 min)    │ (RPO: 1 hr)
  │              │                  │
  │ Active-Active│ CDP-like         │ Snapshot-based
  │ stretched    │ replication      │ to any site
  │ cluster      │                  │ or cloud

NC2 (Nutanix Cloud Clusters) extends DR to public cloud — your on-premises Nutanix cluster replicates workloads to Nutanix running natively on AWS or Azure. Failover runs the same VMs in the cloud with no conversion needed.

Part 6: Network Security — Flow¶

Nutanix Flow Network Security is a built-in microsegmentation solution that requires no additional hardware or agents.

Traditional datacenter security relies on perimeter firewalls — once an attacker is inside, they can move laterally between any two VMs on the same VLAN. Flow solves this with application-centric policies that define exactly which VMs are allowed to talk to which other VMs, on which ports.

Without Flow:
VM-A ──────────────────────────────▶ VM-B (any port, any protocol)
VM-A ──────────────────────────────▶ VM-C (unrestricted)

With Flow policies applied:
VM-A ──── TCP:8080 only ──────────▶ VM-B ✓
VM-A ──✗──────────────────────────▶ VM-C (blocked by policy)

Flow includes a visualization mode that shows actual traffic flows between application tiers before any policy is applied — making it safe to implement least-privilege networking without guessing what will break.

Part 7: Management — Prism and NCM¶

Prism: The Management Plane¶

Nutanix provides two levels of management:

Prism Element runs on each individual cluster. It handles cluster-level operations: VM management, storage configuration, health monitoring, and cluster settings.

Prism Central is a separate VM (or scale-out cluster of VMs) that aggregates all your Prism Element clusters into a single UI. From Prism Central, you manage dozens of clusters as if they were one.

┌──────────────────────────────────────────┐
│             Prism Central                │
│  ┌────────┐  ┌────────┐  ┌────────┐     │
│  │Cluster │  │Cluster │  │Cluster │     │
│  │  NYC   │  │   LA   │  │ Azure  │     │
│  │(PE)    │  │(PE)    │  │(NC2)   │     │
│  └────────┘  └────────┘  └────────┘     │
└──────────────────────────────────────────┘

LCM: Upgrades Without Guesswork¶

Life Cycle Manager (LCM) is Nutanix's automated upgrade system. It handles the full stack:

• AOS (Acropolis Operating System)
• AHV hypervisor
• Prism Element and Prism Central
• Firmware: BIOS, BMC, NIC, HDD/SSD

LCM understands the dependency matrix between all components — it knows that upgrading AOS to version X requires AHV version Y and that certain NIC firmware must be at version Z first. It performs rolling upgrades one node at a time, ensuring zero downtime for running VMs.

NCM: Multi-Cloud Management¶

Nutanix Cloud Manager (NCM) extends beyond infrastructure management to workload management across private and public clouds. It operates across three functions:

BUILD — Self-service cloud provisioning with governance guardrails. Users request VMs, databases, or applications through a catalog; policies enforce quotas, cost limits, and approved configurations.

OPERATE — Intelligent operations including capacity planning, performance analysis, and cost optimization recommendations across all clusters and cloud accounts.

GOVERN — Policy-based governance, rightsizing recommendations, idle resource detection, and chargeback/showback reporting by team, department, or project.

Part 8: Nutanix Kubernetes Platform (NKP)¶

NKP (Nutanix Kubernetes Platform) is enterprise Kubernetes — not just a managed cluster but a full fleet management solution.

┌────────────────────────────────────────────────────────────┐
│                         NKP                                │
│                                                            │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐     │
│  │   Cluster    │  │   Cluster    │  │   Cluster    │     │
│  │  Production  │  │  Staging     │  │  Edge Site   │     │
│  └──────────────┘  └──────────────┘  └──────────────┘     │
│                                                            │
│  Fleet Management │ GitOps (Flux) │ Policy Engine          │
│  CI/CD Integration │ Image Registry │ Observability        │
└────────────────────────────────────────────────────────────┘

NKP provisions Kubernetes clusters on Nutanix HCI, bare metal, or public cloud. It manages the full lifecycle: cluster provisioning, upgrades, day-2 operations, and multi-cluster fleet management via GitOps.

NDK (Nutanix Data Services for Kubernetes) complements NKP with Kubernetes-native backup, disaster recovery, and application-consistent replication — covering stateful workloads (databases, message queues) that standard Kubernetes DR tools often miss.

Part 9: Database Service (NDB)¶

NDB (Nutanix Database Service) is a Database-as-a-Service layer that runs on top of Nutanix HCI. It automates the operational burden of managing heterogeneous databases across hybrid clouds.

Supported databases: SQL Server, Oracle, MySQL, PostgreSQL, MongoDB

What NDB handles: - One-click database provisioning from golden images - Point-in-time recovery and log shipping - Patch management and version upgrades - Cross-site replication for DR - Snapshot-based instant clones for test/dev environments

The instant clone capability is particularly powerful for development workflows: a production-scale database (hundreds of GB) can be cloned to a dev/test environment in seconds, consuming almost no additional storage until the dev data diverges from production.

Part 10: Enterprise AI Infrastructure (NAI)¶

NAI (Nutanix Enterprise AI) is Nutanix's answer to the question: how do enterprises run LLMs and AI workloads on their own infrastructure?

┌───────────────────────────────────────────────────────────┐
│                     NAI Architecture                      │
│                                                           │
│  ┌─────────────────┐   ┌───────────────────────────┐     │
│  │   AI Models     │   │   Secure API Endpoints    │     │
│  │  (NVIDIA NIM)   │   │   (managed, audited)      │     │
│  │  Open Source    │   │                           │     │
│  │  Custom Models  │   │   Rate limiting, auth     │     │
│  └────────┬────────┘   └───────────────────────────┘     │
│           │                                               │
│  ┌────────▼──────────────────────────────────────────┐   │
│  │          Kubernetes (NKP)                         │   │
│  │  GPU nodes │ CPU nodes │ Storage │ Networking     │   │
│  └───────────────────────────────────────────────────┘   │
│                                                           │
│  NVIDIA NeMo integration for fine-tuning                  │
│  Agentic AI workload support (multi-step, tool-using)     │
└───────────────────────────────────────────────────────────┘

NAI deploys on any Kubernetes cluster managed by NKP, on Nutanix HCI. It provides:

• Model deployment: Run NVIDIA NIM-packaged models (Llama, Mistral, Gemma, and others) on GPU nodes with NVIDIA GPU Operator
• Secure API endpoints: Every model gets a managed API endpoint with authentication, rate limiting, and full audit trails — keeping sensitive data on-premises
• Fine-tuning: Integration with NVIDIA NeMo for domain-specific model fine-tuning
• Agentic AI support: Infrastructure designed for multi-step, tool-using AI agents that call external APIs, query databases, and take real-world actions

Part 11: Unified Storage¶

Beyond block storage for VMs, Nutanix provides three additional storage protocols through the Nutanix Unified Storage suite:

Data Lens adds a security and compliance layer for unstructured data managed by File Services and Object Services:

• Ransomware detection: Anomaly detection identifies unusual file modification patterns (common in ransomware attacks) and alerts or blocks access
• Audit trails: Full record of who accessed what files, when, from where
• Data classification: Identify sensitive data (PII, financial records) across file shares

Part 12: Security and Compliance¶

Nutanix takes a layered approach to security across the full stack:

Layer             Security Feature
────────────────────────────────────────────────────────
Platform          FIPS 140-2 validated encryption
                  Common Criteria EAL2+
                  Secure boot, TPM 2.0

Access Control    Role-based access (RBAC)
                  Multi-factor authentication
                  LDAP/AD integration

Network           Flow microsegmentation
                  Encrypted VM traffic (AES-256)
                  Zero Trust network policies

Data              At-rest encryption (AES-256-XTS)
                  In-transit encryption (TLS 1.3)
                  Key management (native or external KMS)

Compliance        FedRAMP Moderate (IL2)
                  DoDIN APL (US DoD use)
                  ISO 27001/27017/27018/27701/28000
                  NIST CSF, DISA STIG
                  NIST Zero Trust Architecture alignment

The Nutanix Cloud Platform Map¶

All products fit into a coherent platform hierarchy:

┌─────────────────────────────────────────────────────────────────┐
│                       Nutanix Central                           │
│         (Prism Central + Cloud Manager + Data Lens)             │
├────────────────────┬────────────────────┬───────────────────────┤
│   NCI              │   NCM              │   NKP                 │
│   Cloud Infra      │   Cloud Manager    │   Kubernetes          │
│   ─────────────    │   ─────────────    │   ──────────          │
│   AOS (storage)    │   Self-Service     │   Fleet Mgmt          │
│   AHV (hypervisor) │   Capacity Ops     │   GitOps              │
│   Flow (security)  │   Cost Govern.     │   NDK (data svc)      │
│   DR & Snapshots   │   Multi-cloud view │                       │
├────────────────────┴────────────────────┴───────────────────────┤
│   Nutanix Unified Storage    │   NDB            │   NAI          │
│   (Block / File / Object)    │   (DB Service)   │   (AI Infra)   │
├──────────────────────────────┴──────────────────┴───────────────┤
│   NC2 — Nutanix on AWS / Azure (hybrid cloud target)            │
└─────────────────────────────────────────────────────────────────┘

Summary¶

Nutanix started with one idea — collapse the datacenter stack into software — and has methodically extended that philosophy upward through every layer of enterprise IT.

Here is what the platform delivers in 2026:

AOS + AHV (NCI) — The foundation. A distributed, software-defined storage and compute platform on commodity hardware. Data locality keeps reads fast. Replication Factor (RF2/RF3) keeps data safe. Erasure coding and dedup reduce storage costs by up to 60%. The built-in AHV hypervisor eliminates VMware licensing costs for 69% of customers today.

Prism + NCM — The management plane. Prism Central gives a single view across every cluster, on-premises and in the cloud. LCM handles full-stack upgrades (software + firmware) rolling and zero-downtime. NCM adds self-service provisioning, intelligent operations, and governance across hybrid environments.

Flow — Microsegmentation without hardware changes. Application-centric security policies that stop lateral movement — the attack pattern that turns a single compromised VM into a full datacenter breach.

NKP + NDK — Enterprise Kubernetes. Not just cluster provisioning but full fleet management with GitOps, CI/CD integration, and Kubernetes-native backup/DR for stateful apps.

NDB — Database-as-a-service. One-click provisioning, instant clones for dev/test, and automated lifecycle management across SQL Server, Oracle, MySQL, PostgreSQL, and MongoDB.

NAI — AI infrastructure for enterprises that cannot send data to a public cloud API. Run LLMs on-premises on GPU nodes, with managed API endpoints, audit trails, and NVIDIA NIM/NeMo integration for fine-tuning.

Unified Storage + Data Lens — Block, file, and object storage unified on the same HCI platform, with ransomware detection, anomaly alerting, and compliance-grade audit trails.

NC2 — The hybrid cloud bridge. Run Nutanix software natively on AWS or Azure for DR, cloud bursting, or datacenter exit — with the same tools, same policies, and same management plane you use on-premises.

The thread connecting all of it: operational simplicity. Every layer Nutanix adds is designed to reduce the number of tools, vendors, teams, and tickets required to keep enterprise infrastructure running. In an era where cloud-native skills are expensive and datacenter expertise is aging out, that simplicity is the product.

Discussion

Have thoughts on this post? Share them below — questions, corrections, or your own experience are all welcome.