karpenter VS keda

Karpenter, is a cluster autoscaling solution that, for the moment, only works with AWS infrastructure, but helps you increase the number of nodes automatically depending on your Pods requirements.

I have a few EKS clusters that are running cluster-autoscaler, and I'd like to replace it with Karpenter. When reading through the documentation this stood out to me:

Karpenter: since we are talking about EKS maybe this kind of autoscaling is worth your time.

Karpenter

This post demonstrates a proof-of-concept implementation that uses Kubernetes to execute code in response to an event here is API request. The workflow is powered by Keda (Kubernetes Event-driven Autoscaling) which scales out the kubernetes pods bases on incoming events such as SQS messages. After keda scaleout pods which are in pending state, Karpenter (Just-in-time Nodes for Any Kubernetes Cluster) bases on provisioners to decide scaleout more nodes

Here is the GitHub repo of the tool, and I wrote an article about it if you're interested.

When searching for a cluster autoscaler for EKS, I found out about Karpenter. It's an open-source Kubernetes native node autoscaler developed by AWS.

💡 Karpenter.sh takes a more intelligent, faster approach to autoscaling and improve availability on AWS.

karpenter Helm Release - Installation of karpenter using Helm chart.

Karpenter is an open-source project by AWS that improves the efficiency and cost of running workloads on Kubernetes clusters (not just EKS, it's designed to have support for multiple cloud providers).

The HPA will use by default only basic metrics like CPU and memory. If you implement your own metrics, you might want to switch the autoscaling to use those metrics instead. To achieve this, the most recommended tool would be to rely on the KEDA controller. This grants you more advanced options for horizontal autoscaling. To use it, you should install it via Helm, and then you can use it by replacing the HorizontalPodAutoscaler object with their ScaledObject custom object.

New Release: KEDA 2.8.2

i highly recommend keda for this. https://keda.sh/

Quite some stuff in the new release: KEDA 2.9.0

This post demonstrates a proof-of-concept implementation that uses Kubernetes to execute code in response to an event here is API request. The workflow is powered by Keda (Kubernetes Event-driven Autoscaling) which scales out the kubernetes pods bases on incoming events such as SQS messages. After keda scaleout pods which are in pending state, Karpenter (Just-in-time Nodes for Any Kubernetes Cluster) bases on provisioners to decide scaleout more nodes

Dynamically scaling your containers is getting easier every day when using something like KEDA. I use Pulsar and there's a Pulsar scaler for it in addition to Kafka, Kinesis, Prometheus metrics, etc.

HPA is more commonly used and is the preferred option for situations with a sudden increase in resource usage. KEDA.sh can help flexibly autoscale the application based on custom metrics, like the size of a queue or number of network requests.

Practically, you'll be replacing stock k8s resources (deployments) with custom ones like Argo Rollouts with Keda autoscaling, so you have to plan the respective Gitops CD pipeline (fluxcd/argocd with some crossplane), as well.

Azure Container Apps is built upon the foundation consisting of open-source technology with CNCF projects like Kubernetes Event-Driven Autoscaling (KEDA), Distributed Application Runtime (Dapr), and Envoy running on the Azure Kubernetes Service (AKS).

keda is Kubernetes Event-driven Autoscaling and it's been used wisely now. In this blog, it provides the way to create Keda scaledobject CRD as code using CDK8S typescript.