Not reinventing the wheel: AWS debuts its own K8s distro, looks to make ML more accessible

AWS re:Invent is here and with it a slew of Amazon product announcements, including container tech, storage, machine learning and its own distribution of container orchestrator Kubernetes. 

Amazon EKS Distro, or EKS-D for short, is the version AWS uses for its managed EKS service and is intended to create reliable and secure Kubernetes clusters.

One might wonder why there’s a need for yet another K8s distro, given the number on the market already. In its version, AWS is promoting project features like extended support, along with a host of launch partners providing integrations and installation help to make it more attractive to enterprise laggards. Having EKS available like this will also interest teams needing an extra setup on-prem, as this can help expunge drift between environments and make using the same tooling everywhere a bit easier.

AWS has also developed a public version of the Amazon elastic container registry. This new variant of the container image registry facilitates – as the name suggests – the public sharing of container images, making them also available to those without an AWS account.

Advertisement

Available for preview is managed application deployment service AWS Proton, which is designed with container and serverless applications in mind and aims to support infra teams to come up with self-service offerings for their counterparts in development. 

Services such as this seem to be quite sought after these days, as they allow developers to access already provisioned resources to use for deployments as required, while giving operations control over what’s available. As such, it’s easier to stay on top of resource utilisation (and the associated costs), as well as enforce compliance and security guidelines.

Proton follows an infrastructure as code methodology for defining application stacks and configuring resources, which means they can be easily versioned and stored in a repository to share and keep track of. The service is also supposed to integrate with all sorts of CI/CD and observability tools and comes at no additional cost (yet).

While containers dominated day one of re:Invent (no, we won’t be going into compute resources here), AWS also introduced users to more machine learning related services and enhancements. Showcase offering SageMaker, for example, now comes with a fully managed repository for machine learning features, so that developers and researchers have an easier way of sharing their model’s attributes and properties across teams. 

SageMaker has also been fitted with a tool to prepare data for the training process. Data Wrangler, as it is called, lets users choose a variety of transformations to “normalize, transform, and combine features without having to write any code”. Additional visualisations are meant to help find inconsistencies and become aware of issues tracing back to faulty data prep quicker. 

Teams that are keen on making use of machine learning but in need of a little more help can see if new portfolio additions of Amazon Lookout for Vision and Amazon Monitron can do the trick. They cover the somewhat typical ML use cases of anomaly detection and predictive maintenance, but are geared towards those looking for a more guided way into the world of practical ML.

- Advertisement -