Tagged “scx_lavd”
-
sched_ext: a sketch of the scx_lavd scheduler (Part 3)
This is the third blog post about sched_ext. It has been a while since the last one, and in the meantime scx_lavd has been deployed to the Steam Frame and to Meta's datacenters. In this post, I sketch scx_lavd — the one idea at its core, inferring which task is latency-critical from how tasks wait and wake each other, and the four things that idea forced on us — the virtual deadline, topology-aware load balancing, core compaction with the energy model, and cgroup bandwidth control as a BPF library.
See all tags.
