Push based load-balancing for fair tasks

on YouTube ↗22m

" summary

Prototype RFC sent after OSPM implements a push-based balancer that replaces global idle masks with a global counter plus a chain of per-LLC nohz masks stored in HD shared objects and traversed from idle CPU context; maintaining per-CPU masks proved too expensive on server workloads. Push currently executes from a busy CPU and adds latency, so offloading via smp_call_function to the idle target is proposed. Experiments show the push approach speeds some cases but “works sometimes and fails equally badly”; a bailout using nr_idle_scans reduces waste from affinity and pileups.

AI-written summary · about

" chapters (8)

1. 0:00Introduction to push load balancing
2. 1:03Preparatory patches for push mechanism
3. 5:00Optimizing SK domain shared assignments
4. 8:37Per-LLC idle CPU masks for push
5. 13:10Idle load balancing trade-offs
6. 16:28Experimental push mechanism results
7. 19:43Bailout mechanism and affinity issues
8. 21:50Handling delayed tasks in load balancing