In a SaaS environment, the compute, storage, and bandwidth resources are often shared among tenants, but this makes it challenging to deduce per tenant cost. A SaaS application running on a Kubernetes cluster on AWS adds a layer of further complexity as far as calculating the per tenant cost. Kubernetes is great at abstracting away the underlying pool of hardware. It almost gives us an illusion of having access to a single large compute resource.

Overview Data transfer costs can play a significant role in determining the overall design of a system. Amazon Elastic Container Registry (Amazon ECR), Amazon Elastic Container Service (Amazon ECS), and Amazon Elastic Kubernetes Service (Amazon EKS) can incur data transfer charges depending on a variety of factors. It can be difficult to visualize what that means […]

Topology Aware Hints (TAH) is a feature that available in Amazon EKS version 1.24. It’s intended to provide a mechanism that attempts to keep traffic closer to its origin within the same AZ on in another location. In this post, we’ll explore how this feature can be used with Amazon EKS, its effects on how traffic is routed between pods within an Amazon EKS cluster when using multiple AZs, and whether this functionality allows Amazon EKS customers to optimize the latency and inter-AZ data transfer costs in this architecture.