For all the numbers showing explosive growth of the public cloud, there remains a stubborn segment of the market where policies and regulations prohibit the moving of data into any kind of public cloud.
The choices for private cloud have been one of two ends of the spectrum: build your own Open Stack implementation or opt for vendor solutions like Microsoft Azure Stack, Oracle Cloud at Customer, and IBM Cloud Private. The do it yourself side has been only for the adventurous, for the IT organizations that built their own compute grids a decade ago.
Emergence of container technology, and with it, the enabling of microservices has fueled the fire. The value propositions are that containers are much lighter weight, and therefore, more nimble than VMs, and with containers, you get the construct that is tailor made for microservices. Now add Kubernetes to the mix, and there has emerged a de facto standard for orchestrating all those containers, which are likely to be delivering microservices.
Containers, microservices, and Kubernetes all map very nicely to the idea of efficient use of cloud resources. At least in theory. Now all you have to think about is configuring and managing all this. Good luck with that.
The challenge with Kubernetes: if you try to implement it yourself, a good description of the experience would be no pain, no gain. And if you want to mount a database or Hadoop platform, until recently you had to go outside of Kubernetes to manage it. It's a bit of a slog, as a recent blog post from Cockroach Labs reads to our mind.
The latest version of Kubernetes addresses that gap with Stateful Sets API that persists the identities of specific Kubernetes "pods." But working with Stateful Sets is not for the faint of heart who must also deal with network resources and storage.
Robin Systems, a startup that has raised in excess of $50 million in Series A funding, is taking matters up a few levels with what it calls a "Hyperconverged Kubernetes platform." Fancy branding aside, Robin's platform is designed to provide single click deployment and automatic failover and recovery environment for database applications deployed in private clouds. Its Kubernetes release is targeted at simplifying deployment by avoiding the need to raw code to the Stateful Sets API.
Robin Systems is middleware that creates application-defined infrastructure. That includes levering Docker containers to isolate the application, but more importantly, applying software-defined storage to make it "application aware." That's the layer that Kubernetes lacks. The smarts of the Robin platform comes at the management piece where all the configurations and commands to deploy and manage failover and recovery are abstracted to a single click. The idea is making your do-it-yourself private cloud have the same automatic deployment that you get from public cloud database as a service PaaS offerings.
Robin Systems is not the first provider to traverse this path. Portworx provides a software-defined storage solution for containerized workloads that supports Kubernetes and other orchestrators such as Marathon, Nomad, and Swarm. It makes storage look and operate like cloud storage. Mesosphere is the original player in this space, which originally supplied the Marathon orchestrator that was part of the original Apache Mesos spec but now also supports Kubernetes.
By being the most recent player to the market, Robin is differentiating itself by zeroing in on Kubernetes and incorporating automation. It touts its Hadoop friendliness with its certification from Hortonworks. Robin Systems wants to automate container deployment for enterprises seeking the best of both worlds for private cloud deployment: avoid the vendor lock-in and avoid at least some of the baggage.
Previous and related coverage
As containers grow more popular, virtual machine deployments are declining.
If you're having trouble deciding between virtual machines or containers, Jack Wallen boils down this choice to two simple bullet points.
A new survey shows C-level executives define hybrid cloud slightly differently than IT specialists; level of cloud adoption also influences views.
How can GPUs and FPGAs help with data-intensive tasks such as operations, analytics, and machine learning, and what are the options?