Disclaimer: Grok generated document.
OpenStack is an open-source cloud computing platform — often described as a "cloud operating system" — that lets you build and manage your own Infrastructure-as-a-Service (IaaS) clouds, similar to how AWS or Azure work under the hood, but completely vendor-neutral and free. It controls large pools of compute, storage, and networking resources across a data center (or multiple data centers), managed via APIs, a web dashboard (Horizon), or CLI tools.
It started in 2010 as a joint project between Rackspace and NASA, and is now governed by the OpenInfra Foundation. As of November 2025, the latest release is 2025.2 "Flamingo" (released October 2025), following the six-month release cycle. The previous one was 2025.1 "Epoxy".
In 2025 it's still very much alive and growing: powering >45 million production compute cores worldwide, with the services market estimated at ~$30 billion and growing 25-32% CAGR through 2030. It's especially strong in telecom (NFV/5G), research/HPC, finance, government, and large enterprises that want to avoid vendor lock-in.
- If you're building cloud-native apps: Many companies run private/hybrid clouds on OpenStack. Understanding it helps you debug why your VMs/containers behave differently than on AWS/GCP.
- DevOps/SRE roles: Deploying, operating, or extending OpenStack clouds is a high-value skill (especially with the VMware-to-OpenStack migrations happening in 2025 due to Broadcom pricing changes).
- Contributing to open source: OpenStack is pure Python (mostly), huge codebase, real-world distributed systems problems. Great for résumé-building.
- Kubernetes integration: Modern OpenStack deployments often run Kubernetes clusters on top (via Magnum or Cluster API providers), so the two worlds overlap a lot.
| Aspect | OpenStack | Kubernetes |
|---|---|---|
| Primary focus | Full IaaS cloud: VMs, bare metal, storage, networking | Container orchestration & application lifecycle |
| Workload type | VMs (primary), containers, bare metal | Containers/pods only |
| Scope | Builds the cloud itself | Runs on top of a cloud (or bare metal) |
| Maturity for VMs | Excellent (Nova hypervisors: KVM, VMware, Hyper-V, etc.) | Limited (requires KubeVirt or similar) |
| Maturity for containers | Good (via Magnum, Qinling, Zun) | Best-in-class |
| Typical user | Telcos, enterprises needing private clouds, HPC | Almost everyone for microservices |
| In 2025 | Strong growth in hybrid/edge, VMware migrations | Dominant for apps, but often runs on OpenStack |
They're complementary, not competitors. The common pattern: OpenStack provides the underlying infrastructure (Neutron networking, Cinder storage), Kubernetes runs the workloads.
OpenStack is extremely modular — there are >50 official projects, but these are the essentials most deployments use:
| Service | Code Name | What it does | Key for engineers |
|---|---|---|---|
| Identity | Keystone | Authentication, authorization, service catalog (RBAC, OAuth, LDAP integration) | You’ll hit this first with tokens |
| Compute | Nova | VM lifecycle: scheduling, creation, resize, live-migration (supports KVM, Hyper-V, VMware, etc.) | The "EC2-like" part |
| Networking | Neutron | SDN: virtual networks, routers, load balancers, FWaaS, VPNaaS (plugins: ML2, Open vSwitch, OVN, SDN controllers) | Complex but powerful |
| Block Storage | Cinder | Persistent block storage for VMs (like EBS; backends: Ceph, LVM, NFS, NetApp, etc.) | Volume attach/detach |
| Object Storage | Swift | Highly available S3-compatible object store (petabyte-scale) | Great for backups, static assets |
| Image Service | Glance | VM image registry (upload, snapshot, format conversion) | QCOW2, RAW, ISO, etc. |
| Dashboard | Horizon | Web UI (nice for demos, but most ops use CLI/API) | Django-based, customizable |
| Orchestration | Heat | AWS CloudFormation-compatible templates (declarative infrastructure) | Infrastructure-as-Code |
| Telemetry | Ceilometer / Aodh / Gnocchi | Metrics, alarms, billing data | Monitoring integration |
Other important ones:
- Magnum: Provisions Kubernetes/ Docker Swarm/ Mesos clusters on OpenStack
- Ironic: Bare-metal provisioning
- Barbican: Secrets/key management
- Designate: DNS-as-a-Service
Users → Horizon (web) / OpenStack CLI / REST APIs
↓ (all requests authenticated by Keystone)
→ Nova (compute) + Neutron (network) + Cinder/Glance/Swift (storage)
↓
Hypervisors (KVM/QEMU most common) + Ceph/LVM for storage + OVN/Open vSwitch for networking
Everything communicates over a message queue (RabbitMQ or Kafka) and uses a database (MySQL/PostgreSQL/MariaDB).
-
Try it locally
DevStack (quick & dirty all-in-one):git clone https://opendev.org/openstack/devstack && ./stack.sh
Or MicroStack / OpenStack-Ansible for something closer to production. -
Deploy a real cluster
- Packaged distributions:
- Ubuntu + OpenStack Charm (Juju) – easiest for beginners
- Red Hat OpenStack Platform (enterprise support)
- Canonical Charmed OpenStack
- OpenStack-Helm (runs on Kubernetes)
- Tools like Kolla-Ansible, OpenStack-Ansible, or TripleO for production-grade.
- Packaged distributions:
-
Code & Contribute
- 100% Python (with some Go/Rust creeping in).
- Use Gerrit for reviews: https://review.opendev.org
- Good first bugs: search "low-hanging-fruit" on Launchpad.
- Join the weekly IRC meetings or the OpenInfra Slack/Discord.
-
Learning Resources
- Official docs: https://docs.openstack.org (start with the Install Guide or Architecture Guide)
- Superuser blog, OpenInfra YouTube channel
- Books: "OpenStack Operations Guide", "Learning OpenStack" (a bit dated but concepts hold)
- Certifications: Certified OpenStack Administrator (COA) is very practical.
Pros
- No vendor lock-in, massive community, runs anywhere
- Extremely customizable (you can plug in any storage/network backend)
- Excellent for large-scale VM workloads, edge, telco NFV
- Growing again thanks to VMware exodus and hybrid-cloud needs
Cons
- Steep learning curve — "configuration explosion" in Neutron especially
- Deployments used to be fragile; modern tools (charms, helm, Airship) have improved this a lot
- Not as "sexy" as Kubernetes for pure app devs (but that's changing with K8s integration)
- Upgrades can be painful (though SLURP releases help)
If you're a software engineer working in infrastructure, DevOps, or cloud platforms, knowing OpenStack is still a solid, high-paying skill in 2025 — especially with the current wave of enterprises moving off VMware. It's not going away; it's evolving into the de-facto open alternative for private/hybrid IaaS, and it plays nicely with the Kubernetes world you probably already know.
Start with DevStack or a small Charmed OpenStack cluster this weekend — you'll spin up your own private cloud in under an hour and instantly understand why people both love and complain about it! 🚀

https://www.openstack.org/
https://www.geeksforgeeks.org/python/aws-vs-openstack/