GigaOm Radar for Leveraging Federated Kubernetesv1.0

1. Summary

This report explores the nascent Federated Kubernetes landscape and looks at current, emerging, and future solutions and approaches that will impact enterprise cloud customers. It also provides insight into the emerging providers in this dynamic market space.

Keep in mind that it is early days still for Federated Kubernetes, both in terms of progress toward an industry standard and the emergence of infrastructure players moving to support this approach (if not the standard itself). Included among these players are the major public cloud providers, such as AWS, Microsoft, and Google.

Also addressed in this report are existing mid-tier players that are a large part of the ecosystem, such as Rancher, CoreOS, and Docker. Finally, there are the third-tier hosted players offering Kubernetes as a service, security systems, and management stacks that all support distributed Kubernetes clusters.

What’s helpful about the Federated Kubernetes approach, whether leveraging the KubeFed standard or a more closed approach, is that the architecture should make it easy to deal with multiple clusters running on distributed systems, even on multiple clouds. This capability is based on using two major building blocks.

• Syncing resources across clusters: As you may expect, syncing resources is the core challenge for those deploying federated or distributed Kubernetes clusters. Mechanisms within Kubernetes can automatically sync deployments on plural clusters, running on local or remote systems, cloud and not cloud.
• Inter-cluster discovery: This addresses the ability to automatically configure DNS servers and load balancers with backends that support all clusters running across many public clouds. This is important considering that while good Kubernetes architects can basically DIY a Kubernetes architecture that’s federated, the ability to automate this architecture in terms of the build and operational processes (DevOp) will hasten adoption.

The benefits of leveraging multi-cloud/Federated Kubernetes include high availability, considering you can replicate active/active clusters across multiple public clouds. Thus, if one has an outage, the other can pick up the processing without missing a beat. Also, you avoid the specter of provider lock-in. Kubernetes acts as the abstraction layer that can remove you from the complexities and native details of each public cloud provider. And in this regard, it replaces both cloud services brokers (CSB) and cloud management platforms (CMP).

Note: In evaluating Federated Kubernetes technology we found that the approaches were all over the place in terms of how clusters are distributed. Most provide a DIY approach, meaning that the building of Federated Kubernetes architecture was hands-on. Indeed, most of the distribution and federation of Kubernetes clusters to date have been built this way, including using many of the tools and technologies evaluated here.