SCHED_BSD vs SCHED_ULE ...

[Robert Watson]

On Sat, 29 May 2004, Marc G. Fournier wrote:

> Is there a circumstance where the older SCHED is better then ULE? 

Yes.  I've noticed differing performance properties based on a number of
factors, including hardware variation and workload. I've also noticed a
couple of bugs in ULE that I'm hoping Jeff Roberson gets a chance to fix
at some point, including some issues with load balancing. 

> Or is the older one something that will eventually just be removed
> altogether?
> 
> If the older does have areas in which it is the better, are there any
> docs comparing the two?

Well, I think it's useful to keep around 4BSD even if only to use as a
performance baseline for understanding where ULE is working better or
worse.  I don't know of any specific documents talking about relative
merits: generally, I would consider if a bug if ULE is consistently slower
in some form.  As such, any document would probably be an excellent bug
report :-).

As people run into performance issues with -CURRENT, the scheduler is
one of the first variables I would ask someone to check, FYI.  And in that
sense, keeping 4BSD around is extremely valuable.

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert at fledge.watson.org      Senior Research Scientist, McAfee Research