What's the fastest way to scale containers on AWS in 2022?
Is EKS faster than ECS? What's faster: serverless workers (Fargate) or to serverful workers (EC2)?
What about App Runner and Lambda?
Now we know: https://www.vladionescu.me/posts/scaling-containers-on-aws-in-2022
Full details are in the blog post, but EKS TL;DR:
- EKS on Fargate is now faster than EKS on EC2 🤯
- EKS on EC2 using IPv6 is a tiny bit faster than IPv4
- EKS on EC2 using Karpenter is a tiny bit faster than Cluster-Autoscaler, but it's complex
https://www.vladionescu.me/posts/scaling-containers-on-aws-in-2022/
Full details are in the blog post, but ECS TL;DR:
- ECS on Fargate improved even more 🚀
- ECS on Fargate using ARM is a tiny bit faster
- ECS on Fargate using Windows containers is surprisingly fast
- ECS on EC2 got much faster with Capacity Providers
https://www.vladionescu.me/posts/scaling-containers-on-aws-in-2022/
Full details are in the blog post, but AppRunner TL;DR:
- super-easy to setup
- on the way to become an awesome service
- don't be fooled by the seemingly small limit at 25 containers! That's a grand total of 50 vCPUs and 100 GBs in a single service!
https://www.vladionescu.me/posts/scaling-containers-on-aws-in-2022/
Full details are in the blog post, but Lambda TL;DR:
- ECS on Fargate can't be directly compared with Lambda!
- Lambda follows the exact published performance metrics
- Lambda with increased limits is ridiculous!
https://www.vladionescu.me/posts/scaling-containers-on-aws-in-2022/
AWS -- somebody else's computer running in a building made using custom concrete, running on custom processors, and where you can instantly get 150 TB of RAM and 90 000 vCPUs!
🤯
Got questions? Want to chat about this? Want to learn more?
Join me and @realadamjkeller tomorrow, April 14 at 19:00UTC / 12pm PST, for a special episode of "Containers from the Couch"!