Amazon EKS Anywhere: Manage your Kubernetes clusters on premises
Allan Naim, Principal Product Manager for Amazon Elastic Kubernetes Service (EKS) talks about managing Kubernetes clusters on-premise.
In this video Allan Naim, Principal Product Manager for Amazon Elastic Kubernetes Service (EKS) talks about managing Kubernetes clusters on-premise.
At 1:04 Allan begins the presentation by explaining how EKS functions at Amazon. EKS is an upstream and certified conformant version of Kubernetes with backported security fixes.
AWS supports four versions of Kubernetes. In addition, EKS makes Kubernetes’ operations, administration, and management very simple. Amazon EKS provides the customers the flexibility to start, run, scale Kubernetes applications on AWS or on-premises.
With Amazon EKS, customers can use Kubernetes for their containerized applications and take advantage of all the performance, scale, reliability, and availability of AWS. At 3:54 Allan mentions the customer challenges. Some of the challenges include leveraging existing investments, governance, complexity of operating Kubernetes across on-premises and cloud.
At 5:33 he talks about the latest announcements from Amazon EKS. EKS Distro offers customers version, patching, and security alignment with Amazon EKS. With EKS Distro, customers can easily rely on the same versions of Kubernetes and its dependencies deployed.
The second announcement is EKS Anywhere. This is a new deployment option that enables customers to easily create and operate Kubernetes on-premises.
Deploying applications through EKS Anywhere
In addition, EKS Anywhere provides an installable software package for operating Kubernetes cluster on-premises and automation tooling for cluster lifecycle support.
Customers will be able to automate the cluster management, reduce support costs and eliminate the redundant effort of having to use multiple tools for operating Kubernetes clusters. EKS Anywhere will be available in the second half of 2021.
At 11:20 Allan talks about one of the key principles behind EKS which is ‘Consistent tooling’. Amazon EKS integration with existing AWS services such as systems manager, app mesh, and monitoring tools allows the extension of these capabilities to on-premises. In addition, it enables customers to centralize management, security, observability across all clusters.
At 13:30, another design principle behind EKS Anywhere ‘GitOps’ is explained. EKS Anywhere will provide an operating model that will enable the customers to use Git as a single source of truth for their desired state of their Kubernetes clusters running across their environments including on-premises and AWS.
At 16:08, it is once again emphasized that the entire EKS operational experience will be Git-driven and consistent across on-premises. A Git controller will run on each cluster that allows customers to version control.
At 16:57 Allan explains that EKS Anywhere will ship with an installer and a command-line interface. A local API will be available to create an opinionated EKS Distro that includes optional defaults consistent with Amazon EKS for container run time, container network interface, storage classes, and observability. By including pre-configured defaults with EKS Anywhere clusters, customers will benefit from reduced-complexity and automate cluster configuration.
In addition, any part of the Kubernetes stack can be customized by bringing any node operating system and service load balancer of choice of ingress. Customers can easily consume AWS services like pulling container images from Amazon ECR, leverage bootstrap application workloads from the container registry. At 18:44 Allan talks about the connectivity options offered by EKS Anywhere. The clusters can either run as fully connected or partially connected.
Connectivity to AWS will be provided by a cluster agent which will securely connect to AWS integrated with AWS Identity provider. At 19:02 it is mentioned that for disconnected scenarios, customers can use the EKS Distro and leverage open source tools to run clusters where there are prolonged periods of disconnection.
At 19:25 Allan elaborates on the steps to be followed to deploy applications with EKS Anywhere. The first step would be to download an installer and a command-line interface. The installer is to create an environment to use the CLI tool to create the first Kubernetes cluster.
Cluster properties such as node size, networking, storage are specified in the provided configuration files. The CLI is then used to create an EKS Anywhere cluster with the required configurations. Once a cluster is provisioned, the cluster life cycle can be managed in version control using Git. AWS now becomes the central place to run and operate the Kubernetes cluster across on-premises.
At 20:44 Allan talks about some of the common use-cases of EKS Anywhere. EKS Anywhere can be used to train ML models in the cloud and running inference in the Edge, application modernization, and data sovereignty. EKS Anywhere helps to transition the stateless parts of the applications to the cloud while keeping the data in place. In this video, the architecture, principles, and benefits of EKS Distro and EKS Anywhere are illustrated.