You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Based on this video: https://www.youtube.com/watch?v=Y4pOmi0j_90, EKS has noticeable network overhead which could be different on different CNI. Cillium for example uses eBPF and replaces kube-proxy which could have faster packet processing.
The text was updated successfully, but these errors were encountered:
Not sure for EKS, but on GKE using Dataplane V2 or AKS using Azure CNI Powered by Cilium would use kube-proxy-free CNI implementation.
It would be interesting to see how different CNIs impact network overhead. While the difference might not be huge, Cilium's use of eBPF instead of iptables should provide a significant performance boost.
If you ever have the time to benchmark it, that would be awesome. Looking forward to your content as always.
Based on this video: https://www.youtube.com/watch?v=Y4pOmi0j_90, EKS has noticeable network overhead which could be different on different CNI. Cillium for example uses eBPF and replaces kube-proxy which could have faster packet processing.
The text was updated successfully, but these errors were encountered: