[gentoo-amd64] PORTAGE_NICENESS

[gentoo-amd64] PORTAGE_NICENESS

[Mark Knecht]

I'm wondering if anyone can recommend a good setting for
PORTAGE_NICENESS that would allow some compiles in the background
without drastically effecting MythTV? Or is there some better way to
do this with other software?

I've fiddled around with renicing mythfrontend but even going to -15
didn't seem to help that much. Seems better to just make the compiles
slower and more friendly.

Note that this seems somewhat worse under the newest -rt kernel -
2.6.21.4-rt12-cfs-v17. I may need to back up and see if the previous
one I was running was any better.

Thanks in advance,
Mark
-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] PORTAGE_NICENESS

[Hemmann, Volker Armin]

On Donnerstag, 14. Juni 2007, Mark Knecht wrote:
> I'm wondering if anyone can recommend a good setting for
> PORTAGE_NICENESS that would allow some compiles in the background
> without drastically effecting MythTV? Or is there some better way to
> do this with other software?
>
> I've fiddled around with renicing mythfrontend but even going to -15
> didn't seem to help that much. Seems better to just make the compiles
> slower and more friendly.
>
> Note that this seems somewhat worse under the newest -rt kernel -
> 2.6.21.4-rt12-cfs-v17. I may need to back up and see if the previous
> one I was running was any better.
>
> Thanks in advance,
> Mark


set it to +19

not only does it free lots of CPU cycles for everybody else, they are 
also 'batch scheduled' which should be good for compiling.

read /usr/src/linux/Documentation/sched-desing.txt

quote:
- batch scheduling. A significant proportion of computing-intensive tasks
   benefit from batch-scheduling, where timeslices are long and processes
   are roundrobin scheduled. The new scheduler does such batch-scheduling
   of the lowest priority tasks - so nice +19 jobs will get
   'batch-scheduled' automatically. With this scheduler, nice +19 jobs are
   in essence SCHED_IDLE, from an interactiveness point of view.
-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] PORTAGE_NICENESS

[Mark Knecht]

On 6/14/07, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de> wrote:
> On Donnerstag, 14. Juni 2007, Mark Knecht wrote:
> > I'm wondering if anyone can recommend a good setting for
> > PORTAGE_NICENESS that would allow some compiles in the background
> > without drastically effecting MythTV? Or is there some better way to
> > do this with other software?
> >
> > I've fiddled around with renicing mythfrontend but even going to -15
> > didn't seem to help that much. Seems better to just make the compiles
> > slower and more friendly.
> >
> > Note that this seems somewhat worse under the newest -rt kernel -
> > 2.6.21.4-rt12-cfs-v17. I may need to back up and see if the previous
> > one I was running was any better.
> >
> > Thanks in advance,
> > Mark
>
>
> set it to +19
>
> not only does it free lots of CPU cycles for everybody else, they are
> also 'batch scheduled' which should be good for compiling.
>
> read /usr/src/linux/Documentation/sched-desing.txt
>
> quote:
> - batch scheduling. A significant proportion of computing-intensive tasks
>    benefit from batch-scheduling, where timeslices are long and processes
>    are roundrobin scheduled. The new scheduler does such batch-scheduling
>    of the lowest priority tasks - so nice +19 jobs will get
>    'batch-scheduled' automatically. With this scheduler, nice +19 jobs are
>    in essence SCHED_IDLE, from an interactiveness point of view.

Thanks Volker.

so do I actually use "+19"?

PORTAGE_NICENESS="+19"

or

PORTAGE_NICENESS=19

Thanks,
Mark
-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] PORTAGE_NICENESS

[Mike Doty]

Mark Knecht wrote:
> On 6/14/07, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>
> wrote:
>> On Donnerstag, 14. Juni 2007, Mark Knecht wrote:
>> > I'm wondering if anyone can recommend a good setting for
>> > PORTAGE_NICENESS that would allow some compiles in the background
>> > without drastically effecting MythTV? Or is there some better way to
>> > do this with other software?
>> >
>> > I've fiddled around with renicing mythfrontend but even going to -15
>> > didn't seem to help that much. Seems better to just make the compiles
>> > slower and more friendly.
>> >
>> > Note that this seems somewhat worse under the newest -rt kernel -
>> > 2.6.21.4-rt12-cfs-v17. I may need to back up and see if the previous
>> > one I was running was any better.
>> >
>> > Thanks in advance,
>> > Mark
>>
>>
>> set it to +19
>>
>> not only does it free lots of CPU cycles for everybody else, they are
>> also 'batch scheduled' which should be good for compiling.
>>
>> read /usr/src/linux/Documentation/sched-desing.txt
>>
>> quote:
>> - batch scheduling. A significant proportion of computing-intensive tasks
>>    benefit from batch-scheduling, where timeslices are long and processes
>>    are roundrobin scheduled. The new scheduler does such batch-scheduling
>>    of the lowest priority tasks - so nice +19 jobs will get
>>    'batch-scheduled' automatically. With this scheduler, nice +19 jobs
>> are
>>    in essence SCHED_IDLE, from an interactiveness point of view.
> 
> Thanks Volker.
> 
> so do I actually use "+19"?
> 
> PORTAGE_NICENESS="+19"
> 
> or
> 
> PORTAGE_NICENESS=19
> 
> Thanks,
> Mark
the latter
-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] PORTAGE_NICENESS

[Richard Freeman]

[-- Attachment #1: Type: text/plain, Size: 1028 bytes --]

Mike Doty wrote:
> Mark Knecht wrote:
>> so do I actually use "+19"?
> the latter

Yes, a -19 would certainly make for a VERY unusable system!

I'm running a myth backend and I've found a few useful tricks:

1.  Myth itself is run at nice -10.  This was accomplished by editing 
the init.d script.

2.  Within myth tasks like commercial flagging are set to run niced. 
You don't want myth spawning random tasks running at -10.  Unfortunately 
I can't find a way to do this with mythfilldb (short of writing a 
wrapper script I suppose).

3.  emerge schedutils and setup your block devices to use a scheduler 
which respects IONICEness.

4.  Run mythbackend using ionice -c 1 for realtime scheduling.

5.  I usually run builds with ionice -c 3 nice -n 19.  That results in 
essentially no impact at all on system performance.

Oh, if you run cron jobs I'd avoid setting them to ionice -c 3.  I had 
problems with backups taking multiple days (instead of an hour) to run 
that way.  I guess it depends what your priorities are...

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3875 bytes --]

Re: [gentoo-amd64] PORTAGE_NICENESS

[Peter Humphrey]

On Thursday 14 June 2007 18:57:13 Hemmann, Volker Armin wrote:

> read /usr/src/linux/Documentation/sched-desing.txt

"design", not "desing"   :-)

> quote:
> - batch scheduling. A significant proportion of computing-intensive tasks
>    benefit from batch-scheduling, where timeslices are long and processes
>    are roundrobin scheduled. The new scheduler does such batch-scheduling
>    of the lowest priority tasks - so nice +19 jobs will get
>    'batch-scheduled' automatically. With this scheduler, nice +19 jobs
> are in essence SCHED_IDLE, from an interactiveness point of view.

That's an interesting description - thanks for the pointer.

The kernel includes three schedulers: anticipatory, deadline and (the 
default) CFQ, whereas the author refers only to "old" and "new" schedulers. 
I wonder which of the features he describes are common to all three. I also 
wonder which kernel version introduced the "new" scheduler. I have a reason 
for this interest, which I intend to describe in a separate thread.

-- 
Rgds
Peter Humphrey
Linux Counter 5290, Aug 93
-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] PORTAGE_NICENESS

[Richard Freeman]

[-- Attachment #1: Type: text/plain, Size: 963 bytes --]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Peter Humphrey wrote:
> 
> The kernel includes three schedulers: anticipatory, deadline and (the 
> default) CFQ, whereas the author refers only to "old" and "new" schedulers. 
> I wonder which of the features he describes are common to all three. I also 
> wonder which kernel version introduced the "new" scheduler. I have a reason 
> for this interest, which I intend to describe in a separate thread.
> 

Those are the IO schedulers.  He was referring to the O(1) process
scheduler.  The former handles block device contention, the latter
handles CPU contention.  Both are important, but these days IO
scheduling is more likely to be the source of slowness...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGdQ62G4/rWKZmVWkRAj+dAJ9IrfodLG3xCywqJnkMJRvZMkgBXQCgjNQK
alBSbxDT3aL6RCLEZZbWBIo=
=Gudr
-----END PGP SIGNATURE-----

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3875 bytes --]

Re: [gentoo-amd64] PORTAGE_NICENESS

[Peter Humphrey]

On Sunday 17 June 2007 11:36:46 Richard Freeman wrote:
> Peter Humphrey wrote:
> > The kernel includes three schedulers: anticipatory, deadline and (the
> > default) CFQ, whereas the author refers only to "old" and "new"
> > schedulers. I wonder which of the features he describes are common to
> > all three. I also wonder which kernel version introduced the "new"
> > scheduler. I have a reason for this interest, which I intend to
> > describe in a separate thread.
>
> Those are the IO schedulers.  He was referring to the O(1) process
> scheduler.  The former handles block device contention, the latter
> handles CPU contention.  Both are important, but these days IO
> scheduling is more likely to be the source of slowness...

Ah. Thanks.

-- 
Rgds
Peter Humphrey
Linux Counter 5290, Aug 93
-- 
gentoo-amd64@gentoo.org mailing list