Re: SCHED_4BSD in RELENG_7 disturbs workflow

From: Nathan Lay <nslay_at_comcast.net>
Date: Wed, 17 Oct 2007 11:35:09 -0400
P.U.Kruppa wrote:
> On Tue, 16 Oct 2007, Kris Kennaway wrote:
>
>> [LoN]Kamikaze wrote:
>>> I know that RELENG_7 is not considered very near-release, but I 
>>> thought I'd
>>> give my 2¢ in the hope that I might have a little influence on the 
>>> scheduler
>>> development to my benefit.
>>>
>>> The switch from RELENG_6 to RELENG_7 went relatively smooth and 
>>> apart from ipw
>>> causing panics. However there is one thing that's disturbing and 
>>> this is the
>>> scheduler. I only have single core machines, so whatever I say only 
>>> applies to
>>> those. If you think single-core machines are no longer important, 
>>> feel free to
>>> ignore this. In deed, just ignore me however much you like.
>>>
>>>> From my perspective scheduling on RELENG_6 was way better. Even on 
>>>> a full
>>> workload like a portupgrade the focused application (both in X and 
>>> on the
>>> console) always received enough cycles to run smoothly and 
>>> applications that
>>> ran in background like audio players also kept on running fine.
>>>
>>> Quite the contrary on RELENG_7. During a portupgrade or even worse 
>>> 'pkgdb -L'
>>> (recovering lost dependencies) audio players (both graphical and 
>>> mplayer)
>>> scatter, either because they don't get the hard-disk or CPU-cycles 
>>> (which one,
>>> I don't know) and the focused application also often hangs. It just 
>>> looks like
>>> occasionally (under load) everything freezes for a second and then 
>>> goes on
>>> relatively normal.
>>>
>>> I've got the impression that things compile a little faster (that 
>>> might be my
>>> imagination, though), but I'd rather have a smooth working experience.
>>>
>>> This is just my view of the situation and I suppose it is only one 
>>> of many. I
>>> bid you be merciful with us single-core people, who cannot afford a 
>>> slick
>>> multi-core machine, because we worry how to pay for our food at the 
>>> end of the
>>> month.
>>
>> Not to say that any problems that might have developed with 
>> SCHED_4BSD should not be fixed, but you should give SCHED_ULE a try 
>> since it brings benefits even for single CPU systems (e.g. better 
>> interactive response).
> I would like to second that. I have seen the same problems on my 
> single processor system and using SCHED_ULE instead of SCHED_4BSD 
> seems to improve the situation a lot.
>
> Uli.
>>
>> Kris
>> _______________________________________________
>> freebsd-current_at_freebsd.org mailing list
>> http://lists.freebsd.org/mailman/listinfo/freebsd-current
>> To unsubscribe, send any mail to 
>> "freebsd-current-unsubscribe_at_freebsd.org"
>>
>
>
>
> Peter Ulrich Kruppa
> Wuppertal
> Germany
> ------------------------------------------------------------------------
>
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
I have this problem too...although I always imagined the cause was 
WITNESS and various other debug options.

Best Regards,
Nathan Lay
Received on Wed Oct 17 2007 - 13:35:46 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:19 UTC