o3k

o3k is a lightweight control plane that speaks OpenStack's APIs without asking you to run a full OpenStack deployment. It packs the core services into a single Go binary of around 35 megabytes, so instead of wiring together a sprawl of components you start one process and point OpenStack tooling at it. The line on the site sums it up, OpenStack-compatible infrastructure without full OpenStack operational weight. It comes from the team behind kubedo.io, and the framing throughout is about evaluation and testing rather than replacing a production cloud.

Anyone who has stood up real OpenStack knows the cost. It's a large distributed system with a database, a message broker, and many services that all have to be installed, configured, and kept alive, which is a lot to carry just to evaluate a pattern, run a lab, or test tooling against a private cloud. o3k trims that down. It implements the core services in-process and talks between them without a broker like RabbitMQ, which collapses most of the moving parts into one binary you can start and stop at will. That single binary comes in at roughly 35 megabytes, small enough to drop onto a laptop or a CI runner without ceremony.

Out of the box it uses an embedded SQLite database, with PostgreSQL available when you want it, and it starts with zero configuration by simply running the binary. It covers the services teams reach for first, identity, compute, networking, image, and block storage, mapped to the familiar Keystone, Nova, Neutron, Glance, and Cinder names. Two modes are on offer, a default stub mode that runs anywhere and is handy for API testing, and a real mode that drives libvirt and KVM with iptables networking for actual virtual machines. The stub mode is the quickest way in, since it runs cross-platform and lets you exercise the APIs without any hypervisor at all, while the real mode is there when you want VMs to actually boot.

The point of aiming for the OpenStack API is that existing tools keep working. o3k answers the standard OpenStack CLI, plugs into Terraform providers, and can serve the Horizon dashboard, so what you learn against it carries over to the real thing. It reports roughly 70 to 80 percent API fidelity per service, and it builds in the operational pieces you'd expect, TLS, JWT authentication, RBAC middleware, and CADF audit logging, plus Prometheus metrics and Grafana dashboards. It also aligns with the Sovereign Cloud Stack standards. Getting started is deliberately blunt, with a one-line install script that fetches the binary, sets up a service, configures networking, and bootstraps some test infrastructure, or a Docker Compose file that can bring up the Horizon dashboard alongside it.

It's aimed at small and medium infrastructure teams weighing a VMware alternative, exploring private cloud and edge patterns, or looking into sovereign cloud approaches. For those groups the appeal isn't running production workloads on it, but getting a fast, faithful-enough sandbox to learn the APIs, script against them, and shake out tooling before committing to the real system. Being honest about where it stands matters here. o3k is a Technical Preview and openly labels itself as such, so it belongs in controlled environments and small-scale labs rather than production. Basic operations work across the services, while pieces like quotas, live migration, and identity federation are still on the gap list, which the project documents plainly rather than papering over. An external review put it around a middling score for lab and small-scale use, and the maintainers keep an honest gap list rather than overselling, which is the right posture for something at this stage. Read that way, the Technical Preview label is a feature, since it sets expectations before you build anything important on top of it.

What sets it apart is the packaging. Rather than a slimmed configuration of the real thing, it's a fresh implementation designed to be small, portable, and quick to stand up, which is exactly what evaluation and continuous-integration testing want. o3k is open source under the Apache 2.0 license, installable with a one-line script, a Docker Compose file, or the raw binary, so there's nothing to buy and nothing stopping you from reading the source. For teams curious about OpenStack without the setup tax, it's a low-commitment way to get hands-on. Because it's a single process with in-process communication rather than a broker-driven fleet of services, the whole thing is easy to reason about, tear down, and rebuild, which is much of what makes it pleasant for repeated testing.