With OpenStack finally making the inroads its proponents have hoped for, the next big questions are where and how to run it, as well as how to keep it fresh.
Several companies have ready answers via turnkey OpenStack solutions: hosted installations in the public cloud, on-premises managed appliances, and mix-and-match solutions that aim to cover most every scenario in between.
Some of the biggest players are OpenStack vendors plying their trade as a service. Others are existing cloud providers offering OpenStack as one of many technologies in their portfolio. Of particular interest is how each handles the business of rolling out upgrades to the OpenStack stack, as OpenStack's components are revised every six months. Here's a look at the heavyweights.
Mirantis OpenStack Express
Mirantis OpenStack Express deserves upfront mention for two reasons. First, Mirantis offers one of the most prominent and easily deployed OpenStack distributions. Second, OpenStack Express is a novel concept: a bare-metal-hosted OpenStack-as-a-service solution, with the bare metal provided by IBM/Softlayer. Among the selling points: No noisy neighbors, thanks to dedicated hosts; straightforward pricing; new editions of the deployed code are kept close to trunk, but tested for stability and reliability; and the user gets direct control over the physical layer as well as OpenStack itself.
It makes sense that one of the original driving forces behind OpenStack would have its own OpenStack turnkey offerings. Rackspace Private Cloud can be used either on its own or in conjunction with Fanatical Support, where Rackspace's team can stand up an OpenStack instance either on its own metal or in a client's data center. Rackspace also makes a point of its OpenStack private cloud leveraging LXC "to independently scale each OpenStack service and enable seamless, in-place upgrades."
Piston's approach to turnkey OpenStack is to provide not just an OpenStack distribution, but also a bottom-up stack -- Piston CloudOS plus Piston OpenStack -- that includes a hardened Linux distribution as a substrate. Among the ease-of-use features professed by Piston is an update service designed to minimize downtime via rolling upgrades and "options for updates to air-gapped installations," for those running a truly private cloud.
A relatively new entry to the field is Platform9 Managed OpenStack, a turnkey solution for private clouds, recently released to the public after a six-month beta test period. Platform9's vaunted features include importing existing ("brown field") environments, no need for OpenStack experience to get started, monitoring and service-level agreements, and having "seamless, nondisruptive updates" pushed out, including once-a-year upgrades to stable versions of OpenStack. (Note: OpenStack is updated twice a year, but not everyone is in the habit of automatically going to the latest edition.)
Few items are more turnkey (albeit more expensive) than a packaged hardware solution for a given software product. Nebula One is a hardware appliance cluster, a central controller plus up to 20 cluster nodes in a given rack, running Nebula's Cosmos edition of OpenStack. Built by former NASA employees who collaborated with Rackspace on the original iterations of OpenStack, it emphasizes not only turnkey deployment, but turnkey management and monitoring as well via an enterprise leveraging its existing monitoring tools. Upgrades, too, are automated.
The overall focus here is less on OpenStack itself and more on its enterprise use: running scale-out workloads well, ramping up or down to meet need, and having it all available in a self-service fashion if needed.
Like OpenStack Express, Ubuntu BootStack is a hosted-on-bare-metal version of OpenStack. Customers can build their own OpenStack clouds on hosted hardware, have Canonical's team tend to the resulting stack (including OS and OpenStack upgrades) with an SLA to match, and even take back full control of their cloud once the prospect of running OpenStack doesn't intimidate you. Various local and remote cloud hardware options are supported, and upgrades to the latest OpenStack editions are part of the deal. Downsides: The costs aren't as transparent as Mirantis' offering, and a three-month commitment is required to get up and running.
A recent Cisco acquisition, Metacloud effectively eclipses Cisco's own OpenStack distribution. Metacloud's engineers perform the setup, provisioning, monitoring, and updating, letting the enterprise concentrate on application development. The resulting OpenStack installation uses an enterprise's existing single-sign-on infrastructure to allow users to provision their own resources, and both OpenStack and AWS APIs can be used for the core services. Unfortunately, as one might imagine for a service this bespoke, pricing is available only on request.
Another big name that's dipped into the OpenStack pond, HP has built a hybrid cloud offering, Helion, on top of its own OpenStack distribution. Both private and public clouds can be built with Helion, with the public cloud offerings hewing closer to Amazon's service and pricing model than anything OpenStack. But rather than offer a single turnkey solution, HP has several managed and consulting services that reflect its existing status as a business solutions provider. Among them: self-service high-performance computing based on OpenStack, a turnkey hybrid cloud system, professional services for OpenStack, and so on.
One possible downside of such a panoply is figuring out what HP offering is the best fit, but HP's Helion OpenStack distribution is always a safe fallback. That distribution was upgraded to the Juno build of OpenStack around the same time Juno itself was released, so it's likely HP's own managed offerings will follow the same closely tracking path.
Aside from economics (including the economics of IT staffing) or manageability, Nebula CEO Gordon Stitt cited another reason why turnkey and managed OpenStack solutions are useful in both on-prem and remote deployments: "The reason to have a hosted OpenStack alternative," he wrote in an email, "is so that enterprises can build a truly seamless hybrid cloud using the same technology on-premise due to security or regulatory requirements and data residency requirements. The residency requirements -- that data never leaves the enterprise -- may be regulatory or may be due to the cost and/or latency of transferring data in and out of a public cloud."