Forward deployment of the AWS Snow family –Architecting for Disconnected Edge Computing Scenarios

Commanders need flexible, rugged compute horsepower that can be quickly deployed or torn down and moved. From an off-the-shelf server’s point of view, forward-deployed Command Posts (CPs) operate under less-than-ideal circumstances. Sometimes, they are semi-permanent structures that have HVAC, but even then, it is limited. CPs rarely meet the environmental requirements of the average rackmount or tower server. Snowball Edge devices meet stringent ruggedization standards, including MIL-STD-810G. This, combined with the ability to use the same APIs and constructs their applications use on AWS in-region, makes them attractive in such situations.

To meet the need for CPU/GPU resources in the CP, customers often deploy multiple Snowball Edge devices. However, Snowball Edge devices themselves do not have a notion of clustering. They do not form a relationship with each other at the infrastructure level the way hyperconverged solutions from Nutanix or Cisco would. While they can freely communicate with each other over the local network, each node is an island of capacity unto itself.

In such circumstances, Amazon Elastic Kubernetes Service Anywhere (EKS-A) can be overlaid on top of multiple AWS Snowball Edge devices to provide these things to a certain degree.

Very similar needs are observed in disaster response scenarios. AWS routinely uses AWS Snow family devices to provide disaster relief to affected communities around the world.

EKS-A

EKS-A is an extension of Amazon’s highly popular managed Kubernetes service that facilitates the deployment, management, and scaling of containerized applications. EKS-A allows customers to run Kubernetes clusters not just in AWS regions but also in their own data centers, other cloud environments, or even in DDIL scenarios.

AWS EKS Distribution

AWS EKS Distribution and EKS-A are both elements of AWS’s broad strategy to provide flexible and robust Kubernetes solutions, but they serve different roles in the ecosystem.

AWS EKS Distribution is the same Kubernetes distribution that is used by all variants of EKS (both in-region and EKS-A). It is freely available on GitHub for anyone to deploy in their environment. The distribution provides a consistent set of binaries vetted by AWS, allowing anyone to set up and manage a Kubernetes environment on their own – with the reassurance that it adheres to AWS’s standards of security and reliability. It does not include integrated tooling for cluster creation and management.

On the other hand, EKS-A is a more comprehensive solution that is built upon the AWS EKS Distribution. It not only includes the core Kubernetes distribution but also provides a fully integrated and automated tooling system for deploying and managing Kubernetes clusters in any environment, be it on-premises or in other clouds.

EKS-A includes things such as an installer and command-line tool that help users create, upgrade, and manage Kubernetes clusters. It also offers an optional connector that can talk back to the EKS service endpoints in-region, allowing customers to visualize the status of remotely deployed EKS-A clusters.