Cloud Services / Kubernetes / Serverless

Why AWS Decided to Let Developers Manage Kubernetes with Serverless

4 Dec 2019 8:47am, by

Following the popularity of its Fargate serverless container offering, Amazon Web Services will now allow users to use Fargate’s serverless model to manage and provision clusters on Elastic Kubernetes Service (EKS). This means that organizations can now specify compute and other criteria for cluster management on EKS as Fargate manages the otherwise complex provisioning and infrastructure management for the Kubernetes pods.

It was one of a number of announcements that AWS CEO Andy Jassy made Tuesday during the company’s annual Re:Invent user conference, being held in Las Vegas this week, suggesting how aggressively AWS is broadening its offerings and tools for use on its cloud infrastructure. Other new offerings Jassy disclosed during his three-hour keynote included new database technologies and AWS’ integrated development environment (IDE) SageMaker Studio for machine learning that the company has launched for the show.

Claiming more than 80% of all containers on the cloud run on AWS while 40% of new container customers opted for serverless container management with Fargate, AWS saw the obvious need to extend Fargate to provision and manage Kubernetes with EKS, Jassy said. The concept of opting to use Fargate for Kubernetes is straightforward: You specify the “CPU and memory you want and upload the container image, and Fargate does all the rest for you” without having to manage servers or provision Kubernetes clusters, Jassy said.

“We did not anticipate Fargate being this popular — customers said, understandably, ‘we love the idea of Fargate, but why won’t you make it work for Kubernetes’?” Jassy said. “Now, our Kubernetes customers are able to get all of the serverless benefits of running containers.”

The option of being able to rely on Fargate to use a serverless platform to provision and manage Kubernetes clusters should resonate well among developers, especially those who work in resource-limited organizations, Torsten Volk, an analyst for Enterprise Management Associates (EMA), said. “Sizing, deploying, configuring, and operating Kubernetes clusters on top standard EC2 infrastructure requires skill, experience, and time and it always leads to significant over-provisioning, just to be on the safe side,” Volk said. “Fargate takes all this away and with EKS on Fargate developers can simply define their application requirements in the form of a few CLI commands. Fargate then goes out and provides the required Kubernetes infrastructure to run their application PODs without the need for code changes.”

Since users can configure Fargate to automatically run pods that are designated to a specific namespace, they have access to “Kubernetes-as-Code,” without the need to worry about cluster updates, configuration errors, or over-provisioning, Volk said. “EKS on Fargate is the container equivalent of the traditional VM vending machine,” Volk said.

Besides removing the need to manage multitenancy, Fargate for EKS, the serverless model, allows users to forgo having to patch, scale or secure a cluster of Amazon EC2 instances, Martin Beeby, a principal evangelist for AWS, wrote today in a blog post on how to provision Kubernetes clusters with Fargate.

“Using Fargate, customers define and pay for resources at the pod-level,” Beeby wrote. “This makes it easy to right-size resource utilization for each application and allow customers to clearly see the cost of each pod.”

The new service should help financial services company Square to save resources when managing its services designed to help sellers run their businesses from secure credit card processing to point of sale solutions, Geoff Flarity, engineering manager for Square’s CashApp, said.

“As we modernize our stack with EKS, we are always looking for opportunities to increase our security posture and lessen our administrative burden,” Flarity said. “The ability to right-size portions of our compute consumption, ensuring optimal utilization without having to spend cycles on capacity planning or operational overhead, is extremely compelling. This is without a doubt the most exciting Kubernetes announcement of the year.”

Datadog’s SaaS-based data analytics platform should also benefit, Ilan Rabinovitch, vice president of product management, Datadog, said. “Containers and orchestration are becoming a standard practice for organizations looking to operate efficiently at scale,” Rabinovitch said. “We’ve seen wide adoption of AWS Fargate throughout our customers. We are excited to see support extend to cover Amazon EKS, so that our customers can further simplify management of Kubernetes at scale on AWS.”