EEVDF Scheduler May Be Ready For Landing With Linux 6.6

[phoronix]

Phoronix: EEVDF Scheduler May Be Ready For Landing With Linux 6.6

Intel Linux engineer Peter Zijlstra's EEVDF CPU scheduler code to replace the existing Completely Fair Scheduler "CFS" code looks like it will attempt to land with the upcoming Linux 6.6 merge window...

https://www.phoronix.com/news/Linux-6.6-EEVDF-Likely

[timofonic]

Benchmarks please!

[Linuxxx]

Hey CFS supporters, where y'all at??

(Imagine what kind of an autist one has to be to feel attached towards a CPU scheduler...)

Anyhow, EEVDF landing into this year's LTS Linux kernel will benefit Ubuntu 24.04 LTS and all the other distributions based upon it, so expect Linux on the desktop to become much more fluid OOTB.

[damentz]

Benchmarks please!

Yes, there's a severe lack of benchmarks comparing EEVDF to the incumbent. Not only that, scheduler refinement takes years to get all the edge cases that only become obvious once a scheduler is mainlined.

Maybe we need a comprehensive front page benchmark to figure out what type of wins and losses we'll actually see across the board?

[Quackdoc]

doesn't this seem a bit rushed?

[timofonic]

Yes, there's a severe lack of benchmarks comparing EEVDF to the incumbent. Not only that, scheduler refinement takes years to get all the edge cases that only become obvious once a scheduler is mainlined.

Maybe we need a comprehensive front page benchmark to figure out what type of wins and losses we'll actually see across the board?

I would donate Michael to do that if I had money, really...

[EphemeralEft]

Contradicting existing paradigms is fine, but it’s arrogant to try and kill the competition because you think your solution is so much better.

Even if the performance is better in every case, there’s no way it’ll be as reliable or secure for at least the first several months of general availability.

[V1tol]

A very biased opinion incoming.

EEVDF is better than other "CFS killers" like PDS or BMQ is that games/apps don't hang under heavy CPU usage with it.

Also I use my computer to do work and EEVDF scheduler results in better p99 latencies during my server app loadtest on localhost and slightly better RPS.

No other changes noticed.

[hyperchaotic]

Benchmarks please!

I wonder how they test its performance, the use cases are complex. For example it probably wont give a game more FPS but it might improve upon stutter/jitter or your GUI might be more responsive when the system is under load. Well, unless it changes how cache is hit or something like that.

[Mitch]

A very biased opinion incoming.

EEVDF is better than other "CFS killers" like PDS or BMQ is that games/apps don't hang under heavy CPU usage with it.

Also I use my computer to do work and EEVDF scheduler results in better p99 latencies during my server app loadtest on localhost and slightly better RPS.

No other changes noticed.

I'm very interested in seeing how these benefits affect various things. Does this mean that say, a webserver that's light loaded or heavy loaded may service its many tasks better and quicker than otherwise? Would audio latency possibly be better under some conditions? Maybe stuttering or frame pacing in videogames?

Not biased for any particular scheduler. I think results for any scheduler or change speak for themselves. This latency improvement is just fascinating to me, and I'm trying to understand where those benefits are best noticed in real world conditions. All around exciting nonetheless