Evaluate eBPF and how it can be applied (or not)

[JorTurFer]

Proposal

Currently, we are deploying the interceptor instances as containers, which works well, but eBPF could improve the performance working at kernel level instead of at application level.

Using eBPF we should be able to get the metrics about the traffic, so technically we should be able to get all the metrics that we need for scaling but we still have to check if we can "hold" the request during cold starts.

Actually, this issue is just a place to bump thoughts to plan next steps in this direction than a real issue or feature request. We should define the next steps from here.

[isala404]

Hi,
We are using the KEDA HTTP add-on to provide scale-to-zero functionality for our internal platform, and one day, this exact thought popped into my mind. So, I just made a really quick POC using Rust and Aya, and the results were promising.

https://github.com/MrSupiri/scale-to-zero-ebpf

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.