cloudparts

OpenStack

Open-source software for running your own infrastructure-as-a-service cloud: a set of services with REST APIs that hand out virtual machines, networks, disks and images on your own hardware, released every six months.

OpenInfra · Confirmed
What it is
Infrastructure as a service (IaaS)
Used for
Virtual machines
Type
Open source
Made by
OpenInfra Foundation (part of the Linux Foundation since 2025)
Foundation
Open Infrastructure Foundation, confirmed, joined September 2012. Part of the Linux Foundation
License
Apache 2.0
Facts checked
13 October 2026
Website Source code

What it is

OpenStack is a set of independent services that each own one resource and expose it through a REST API. A minimal cloud installs, in this order, Keystone (identity, which every service authenticates through), Glance (images), Placement (tracks what each host can offer), Nova (compute) and Neutron (networking); Horizon (the web dashboard) and Cinder (block storage) are the next ones to add, and dozens more cover object storage, DNS, bare metal and so on. Each service has an API process plus worker processes that talk over an AMQP broker such as RabbitMQ and keep their state in a database such as MySQL or MariaDB. Nova drives KVM, QEMU (generally only for development), LXC, VMware vSphere, Virtuozzo and z/VM, and Ironic for bare metal; most installations use only one.

Users get the same self-service model as a public cloud: projects, quotas, flavors, images, private networks, floating IPs and security groups, through the dashboard, the openstack CLI, SDKs or plain HTTP. Operators pay for that with real work, since the install guide’s example needs at least two nodes and is explicitly not for production; real clouds use deployment projects built on tools such as Ansible or Puppet. A coordinated release ships every six months, named by year: 2026.2 Hibiscus came out on 30 September 2026 with AMD SEV-SNP and Intel TDX confidential instances in Nova. Every other release is a SLURP release, so operators can upgrade once a year instead of every six months.

Why teams pick it

  • A full private IaaS under the Apache 2.0 license: compute, networking, block and object storage, identity and images, with no license fees
  • Every resource is behind a documented REST API with a common CLI and SDKs, and the dashboard, the CLI, the SDKs and plain curl all go through the same calls
  • Pluggable underneath: several hypervisors in Nova, a choice of message broker and database, and Ironic for bare-metal machines through the same API
  • Predictable cadence: a release every six months, and SLURP releases that support upgrading once a year (for example 2025.1 Epoxy to 2026.1 Gazpacho)

Watch out for

  • Many moving parts: even the minimal cloud is five services plus a message broker and databases, and the official install guide is a learning setup, not a production one
  • Upgrades remain a project: only adjacent releases and SLURP-to-SLURP jumps are supported, and a maintained series becomes unmaintained or end of life after about 18 months
  • You own the hardware, capacity planning and on-call; there is no managed control plane unless you buy one from a distribution or hosting provider
  • Security fixes need patching across many services: the 2026.2 Hibiscus highlights list cross-tenant fixes in Designate (DNS) and Cyborg (accelerators)

Alternatives

What all infrastructure as a service (iaas) have in common →Compare side by side →
Cherry ServersCommercialA Lithuanian host that rents dedicated bare-metal servers through an API, billed by the hour, as spot capacity or by the month, from seven locations in Europe, the US and Asia.HetznerCommercialA German host that rents cloud VMs by the hour and whole dedicated servers by the month from its own data center parks in Germany and Finland, plus cloud-only sites in the US and Singapore.OVHcloudCommercialA French cloud provider that builds its own servers and data centers and sells bare-metal servers, an OpenStack-based public cloud and hosted private clouds.ScalewayCommercialThe cloud arm of France’s iliad group: VMs, GPUs, bare-metal servers, Kubernetes and managed databases in four three-zone regions in Paris, Amsterdam, Warsaw and Milan.

How it fits with other technologies

Launch a VM and give it a public address with the openstack CLI · bash

# clouds.yaml holds the Keystone URL, project and credentials for mycloud
openstack --os-cloud mycloud flavor list
openstack --os-cloud mycloud image list
openstack --os-cloud mycloud server create --flavor m1.small --image ubuntu-24.04 \
  --network private --security-group default --key-name mykey web1

openstack --os-cloud mycloud floating ip create public    # allocate from the external network
openstack --os-cloud mycloud server add floating ip web1 203.0.113.10
openstack --os-cloud mycloud server show web1 -c status -c addresses

See it in a real stack

No architecture uses OpenStack yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2010–2026

On 19 July 2010 Rackspace and NASA announced OpenStack as an open-source cloud platform: Rackspace released the code behind its Cloud Files storage service under the Apache 2.0 license, and NASA contributed the compute engine of its Nebula cloud platform. Object Storage (Swift) came first, and Compute (Nova), combining Nebula with Rackspace’s Cloud Servers technology, was due later that year.

The first release, Austin, shipped Nova and Swift on 21 October 2010, and a new release has followed every six months since, named alphabetically (Austin, Bexar and so on) until Zed in 2022 and by year from 2023.1 Antelope. The project grew to dozens of deliverables per release, and today the Open Infrastructure Foundation hosts it alongside Kata Containers, StarlingX and Zuul.

As clouds grew, upgrading every six months became a burden, so in February 2022 the Technical Committee started SLURP releases, which allow one upgrade per year. In March 2025 the OpenInfra Foundation announced it would join the Linux Foundation, closing the deal in June 2025, with OpenStack named alongside Linux and Kubernetes in their Open Infrastructure Blueprint.

Why it’s built this way

Open source under Apache 2.0
Rackspace said it founded the project to help drive industry standards and avoid vendor lock-in, so anyone could run the same cloud software on their own hardware.
Separate services with public APIs
Each service owns one kind of resource and talks to the others only through their public APIs, with Keystone as the shared identity service, so operators can deploy, scale and replace them independently.
A fixed six-month release cycle
Projects that take part in the coordinated release ship together, and upgrades were tested and supported between adjacent coordinated releases.
SLURP instead of a slower cycle
The Technical Committee found no consensus on a one-year or 18-month cycle and judged double-digit months between landing a feature and using it in production too long, so it kept six months and let operators skip every other release.

Timeline

  1. 2010

    Rackspace and NASA announce OpenStack on 19 July

  2. 2010

    First release, Austin, with Nova and Swift on 21 October

  3. 2022

    Technical Committee adopts SLURP releases for yearly upgrades

  4. 2023

    Year-based release names start with 2023.1 Antelope

  5. 2025

    OpenInfra Foundation joins the Linux Foundation in July

  6. 2026

    2026.2 Hibiscus released on 30 September

Sources13 · facts checked 13 October 2026

Improve this page on GitHub