TechTarget | December 8, 2016

Enterprises get first glimpse into Kubernetes for Windows

Kubernetes enjoys momentum in the enterprise market, which will be bolstered by support for more operating systems and programming languages.

 Enterprise Kubernetes users can put the platform through its paces in Windows environments, as the public preview of Kubernetes 1.5 becomes available early next week.

Windows Server 2016 and Windows Containers will be supported by the container orchestration platform some large enterprises now run in production. Kubernetes 1.5 — the result of a collaboration between Microsoft, Google, Apprenda and Red Hat — also supports both Windows Containers in Windows Server 2016 and Hyper-V Containers, as well as Windows app development tech, such as Internet Information Server, ASP.NET and .NET Core.

Current users welcomed the news that they can run Kubernetes for Windows deployments now, as well as other OSes.

“We leverage a broad spectrum of operating systems for our application workloads,” said Justin Erenkrantz, head of compute architecture for Bloomberg, the global finance, media and tech company, based in New York. “This is a testament to the inclusive and open nature of the Kubernetes ecosystem.”

Kubernetes in production gains momentum

Erenkrantz oversees an environment that runs both OpenStack and Kubernetes in production in support of the Bloomberg website. He said his shop has experimented with nearly every kind of container orchestration system.

Over the last 18 months, there has been an effort to make Kubernetes highly available and reliable. One example is the added multiregion deployment support. Erenkrantz said, for him, it was a key differentiator for Kubernetes.

“We’re providing financial data to the capital markets, and it’s very important for us to have a highly available and reliable system,” he said.

Kubernetes will soon be put into production at a large financial services company as well, after Docker Swarm didn’t survive tests at scale.

“Docker doesn’t have all of the capabilities you need to really manage the applications once you start to do auto-scaling and you have systems that are more complex than just a single microservice,” said a senior vice president at the financial services company, who declined to be named.

However, before Kubernetes in production can become a reality, the SVP’s company must decide which distribution of Kubernetes to adopt — a tricky proposition, as each vendor has certain “biases” in their implementation of the platform.

Red Hat, which the financial services company relies on heavily for Linux operating systems, has the OpenShift managed container application platform, which coincidentally broadcasted the general availability of an OpenShift Dedicated service on Google Cloud Platform this week as well, at the same time that Kubernetes for Windows integration debuts.

“We’d be more comfortable with Red Hat if they didn’t wrap all the OpenShift stuff around it,” the SVP said. “We almost have to peel things off to make use of it.”

Container market confusion to intensify in 2017

Despite the increased instances of enterprise Kubernetes in production and the new support of Kubernetes for Windows, most companies are slowly ramping up to containers. Their work is challenging, as they must evaluate many choices in container orchestration — from multiple distributions of Kubernetes to a plethora of alternative platforms.

Container orchestration vendors acknowledge this.

“Because the market is so dynamic, a lot of the early majority is sitting back and waiting for there to be calmer seas,” said Armon Dadgar, co-founder and CTO of HashiCorp, which markets the Terraform hybrid cloud platform as well as the Nomad cluster scheduler and Vault secrets management software.*

However, most enterprises are still containerizing one app at a time versus jumping directly from VM-based infrastructures to complex orchestrated environments. Also, most enterprise architectures are not monolithic to begin with and are becoming evermore multilayered in terms of app deployment methods.

“It’s too early to say, ‘What is the right model, what is the line between what I run in bare metal, in a virtual machine, in a container and serverless?'” Erenkrantz said.