Closed beta General availability Q1 2027. Request beta access →
IZT Cloud CLOUD FABRIC
See pricing Start free
Comparison

Fabric vs Harvester

Harvester runs virtual machines as Kubernetes objects on top of a container orchestrator. That is exactly right if your world is Kubernetes. Fabric is built the other way round: a VM-first appliance that can hand storage to Kubernetes later, with an operator surface that never asks you to learn the internals.

Comparing Fabric with Harvester HCI (SUSE Virtualization) (SUSE). Updated 2026-09-03. Fabric is in closed beta today. General availability is planned for Q1 2027. Competitor facts are drawn from public documentation and published or verified pricing; corrections welcome.

The short version
  • Harvester is free and open source, supported through SUSE subscriptions; Fabric is free in Community and per node by class when supported.
  • Harvester’s storage, networking and HA are Kubernetes primitives; Fabric’s are appliance features — storage policies, per-tenant SDN, verified fencing.
  • Harvester wins for platform teams standardised on Rancher; Fabric wins for infrastructure teams who need a tenant portal, a 24×7 SLA and fleet-wide atomic updates.
Pricing at a glance

One dual-socket server, 32 physical cores, fully loaded with VMs (annual, list unless noted)

Fabric — Medium class
$1,800/ node / yr

Fabric Community is free and unlimited; Medium class $1,800 per node per year with business-hours support, 24×7 at +30%.

Harvester
Free software; support priced by SUSE subscription

Model: Free and open source (Apache 2.0). Enterprise support through SUSE subscriptions, typically alongside Rancher Prime.

Basis: public SUSE pricing is quote-based, 2026.

Dimension by dimension

Fabric vs Harvester, honestly scored.

Where Harvester is ahead we say so. Roadmap items are labelled and count against Fabric until they ship.

6 Fabric 2 Harvester 4 Even 12 dimensions · roadmap items count against Fabric until they ship
Fabric compared with Harvester, dimension by dimension
DimensionFabricHarvesterEdge
Licensing model Per physical node by node class (five core bands: ≤8 / 9–16 / 17–32 / 33–64 / 65+), every feature in every class, free unlimited Community edition. No per-core meter, no feature matrix.Open source, free; enterprise support via SUSE subscription, quote-based.Even
What you install A signed firmware image per node plus a control-plane appliance; the operator never sees a container orchestrator.An immutable OS image running a Kubernetes distribution, with VMs as custom resources; operated through Rancher.Even
Storage Distributed block storage pooled across every node, driven by storage policies: ordered device tiers, copies and minimum copies, host- or rack-level failure domains, per-disk IOPS/bandwidth QoS and inline compression with in-place re-pack. No SAN and no storage licence.Replicated block volumes managed by the cluster; per-volume replica count, no tiering policy or failure-domain model across racks.Fabric
Networking Per-tenant SDN in the platform: overlay or VLAN-backed virtual networks, dual-stack IPv4/IPv6, built-in DHCP/DNS, micro-segmentation policies, distributed stateless gateways and ARP or BGP announcement of public prefixes. Not a separate SDN licence.VLAN networks and cluster networking via the container network layer; tenant DHCP, gateways and micro-segmentation are Kubernetes policy work.Fabric
High availability Acknowledged 10-second heartbeats, a liveness verdict in about 20 seconds, and no restart elsewhere until the failed node is verifiably fenced — powered off through its own BMC and blocked at the storage layer.VM rescheduling when a node is declared not ready, relying on volume replicas; no out-of-band power fencing.Fabric
Live migration Live migration; hibernate and resume on any node.Live migration supported.Even
Updates and lifecycle One signed A/B firmware image per node carries the hypervisor, storage, networking and agent. A rolling update evacuates each node by live migration, flips its image slot and rolls back automatically if it does not come up healthy.Cluster-wide upgrades of the OS and orchestrator that must be planned carefully on small clusters.Fabric
Multitenancy Built in. Tenants, quotas, isolated per-tenant networks with their own DHCP/DNS, and a tenant self-service portal with browser consoles — the same product an MSP bills monthly per node.Namespaces and Rancher projects with RBAC — good for platform teams, not a tenant self-service portal.Fabric
Snapshots and backupRoadmap Snapshots and cross-site replication are on the roadmap.Volume snapshots and VM backups to S3/NFS built in.Harvester
Kubernetes integrationRoadmap A CSI driver exposing storage policies to Kubernetes is on the roadmap; VMs and containers side by side.Native: Harvester is Kubernetes, with guest clusters provisioned through Rancher.Harvester
Air-gapped operation Designed for it: local sign-in (password, TOTP, passkeys — never an emailed link), a control plane on your own appliance, signed updates imported and verified on site, no phone-home, no kill-switch.Air-gapped installs are supported with a private registry.Even
Hardware Commodity x86-64, any drive mix.Commodity x86-64; heavier baseline resource footprint per node.Fabric
Who should pick which

Choose Fabric if… Choose Harvester if…

Fabric is the better fit for

  • Infrastructure teams who run VMs first and want an appliance, not an orchestrator
  • MSPs and internal clouds that need a tenant portal
  • Sites that need verified fencing and rollback-safe fleet updates
  • Buyers who want a vendor 24×7 SLA per node

Harvester is the better fit for

  • Platform teams standardised on Rancher and Kubernetes
  • Shops that want VMs and containers under one Kubernetes API today
  • Open-source-first organisations with SUSE relationships

Moving from Harvester to Fabric

Harvester VMs export as disk images and import into Fabric as volumes. Teams that keep Kubernetes as the application layer can later consume Fabric storage policies through the planned CSI driver.

See migration use cases
FAQ

Fabric vs Harvester: common questions

No. Fabric is a VM-first appliance: a signed node image, a control-plane appliance and a REST API. A CSI driver to serve Fabric storage to Kubernetes clusters is on the roadmap.

More comparisons

How Fabric stacks up against the rest.

Fabric vs VMware VMware vSphere and VMware Cloud Foundation (VCF)Fabric vs Proxmox Proxmox Virtual Environment (Proxmox VE)Fabric vs Nutanix Nutanix Cloud Infrastructure (NCI) with the AHV hypervisorFabric vs Scale Computing Scale Computing SC//Platform (HyperCore)Fabric vs VergeIO VergeOS by VergeIOFabric vs Azure Local Microsoft Hyper-V and Azure Local (formerly Azure Stack HCI)Fabric vs XCP-ng XCP-ng with Xen Orchestra (Vates) All comparisons → Methodology and the full list

Try the whole platform free.

Community is unlimited nodes, every feature, forever. Move to a paid class when you want the production update channel and a support contract.