Ellis Tarn (@ellistarn) profile picture

Software engineer @ AWS EKS, Created #kfserving #karpenter, social media illiterate

Seattle, WA
Joined December 2013
Photo by Ellis Tarn (@ellistarn)Photo by Ellis Tarn (@ellistarn)
And there we are. Can't wait to see the future of the project and more implementations to let other cluster owners take advantage of it!
It's live! Congrats team! 💖💕✨💐🍰🧁
1
3
2
15
1,596
Efficient node scaling is critical for making cloud native cost effective. #Karpenter is a key step to making scaling more optimized and dynamic in #Kubernetes and we are happy to announce the AKS Karpenter Provider github.com/Azure/karpenter #kubecon Kudos to the AWS team for starting this innovative project. We look forward to continued partnership on it under the SIG-Autoscaling umbrella.
5
56
13
142
27,317
Ellis Tarn retweeted
Nathan Taber, Head of Product - EKS announcing donation of Karpenter project to SIG-Autoscaling in Kubernetes during the "Reduce Reuse Recycle" keynote! #kubernetes #karpenter @kubernetesio @CloudNativeFdn
27
4
124
33,756
I gave a talk at Kubecon EU in 2021 that discussed a design for complementary optimization loops for node Provisioning and Consolidation. I spent a year clarifying to users that it our vision, but not yet a reality. Now it's finally here! karpenter.sh/v0.16.1/tasks/d…
1
Ellis Tarn retweeted
I’m decided move our EKS clusters to Karpenter from Cluster Autoscaler after @rothgar’s Karpenter demo at AWS Community Day Turkey. We have been using Karpenter at 7 EKS cluster at the moment and It works like a charm! Thanks #AWS for such an awesome scaling solution!
4
8
Creating Tenant Node Pools with Karpenter disq.us/t/44k9k8b
We'd love to collaborate on a GKE Cloud Provider for Karpenter.
This quoted post is unavailable.
AWS is launching Karpenter, an open source autoscaler for Kubernetes clusters tcrn.ch/31hzXgX by @ron_miller
1
12
2
26
Ellis Tarn retweeted
Have you been wondering how to continuously train and deploy your models? This post describes how we solved this problem 👏 without forcing our data scientists to give up @ProjectJupyter 👏 and still using common #devops patterns like #gitops. blog.kubeflow.org/mlops/ #mlops
Ellis Tarn retweeted
We've released Kubeflow 0.7, with beta functionality for many components in preparation for Kubeflow 1.0 & exciting additions like #kfserving 0.2. This was a huge group effort. Thanks to all who contributed! medium.com/kubeflow/kubeflow…