FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Redhat > Fedora Development

 
 
LinkBack Thread Tools
 
Old 01-01-2008, 11:48 PM
Douglas McClendon
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Can someone tell me if there is anything I can do to de-regress f8's
pm-hibernate?

What I mean is that it seems like in f7, pm-hibernate saved the kernel
buffer/filesystem caches as well as the minimal needed stuff. Now in
f8, every time I resume, system performance is crap because everything
needs to be read piecemeal from disk into the caches again.

My analysis is pure speculation based on my understanding of how you can
tune that behaviour with suspend2(tux-on-ice). But it is very noticable
and very annoying and very clearly a f7->f8 change.

-dmc

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-02-2008, 04:53 AM
Douglas McClendon
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Till Maas wrote:

On Mi Januar 2 2008, Douglas McClendon wrote:

Can someone tell me if there is anything I can do to de-regress f8's
pm-hibernate?


pm-utils hibernation code does not differ much between f7 and f8.


My analysis is pure speculation based on my understanding of how you can
tune that behaviour with suspend2(tux-on-ice). But it is very noticable
and very annoying and very clearly a f7->f8 change.


I guess it is then a change in the kernel, but afaik the f7 and f8 kernel are
very similiar, too.


Again, mainly I posted this to see if anyone else is noticing the same
thing.


Basic behavior: I keep several main apps open in multiple desktops,
often hotkey switching between desktops. First thing I do on resume is
often cycle through my desktops (thunderbird, firefox, gnome-terminal,
rhythmbox....).


In F7, cycling would be as responsive after resume as immediately prior.
Now with F8, it thrashes reading from disk for a couple seconds
between each desktop/application switch. Very gross.


And suspend2(tux-on-ice) does have an explicit config variable you can
tune, if you actually want to optimize for hibernation speed rather than
post-resume performance. So my reaction was that the non-user-tweakable
version in fedora's suspend1 must have changed from f7-f8. Yes, kernel.
But I figure I'd gripe here since I was in the mood to gripe. I'm
still hoping I can get someone else here to confirm/deny the significant
change in behavior, and better yet explain, and better yet, explain how
to revert.


-dmc

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-02-2008, 09:59 AM
Till Maas
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

On Mi Januar 2 2008, Douglas McClendon wrote:
> Can someone tell me if there is anything I can do to de-regress f8's
> pm-hibernate?

pm-utils hibernation code does not differ much between f7 and f8.

> My analysis is pure speculation based on my understanding of how you can
> tune that behaviour with suspend2(tux-on-ice). But it is very noticable
> and very annoying and very clearly a f7->f8 change.

I guess it is then a change in the kernel, but afaik the f7 and f8 kernel are
very similiar, too.

Regards,
Till
--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-02-2008, 11:17 PM
Douglas McClendon
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Till Maas wrote:

On Mi Januar 2 2008, Douglas McClendon wrote:

Can someone tell me if there is anything I can do to de-regress f8's
pm-hibernate?


pm-utils hibernation code does not differ much between f7 and f8.


My analysis is pure speculation based on my understanding of how you can
tune that behaviour with suspend2(tux-on-ice). But it is very noticable
and very annoying and very clearly a f7->f8 change.


I guess it is then a change in the kernel, but afaik the f7 and f8 kernel are
very similiar, too.


http://lwn.net/Articles/153203/

Just for the benefit of future web searchers with the same question,
here is an interesting thread covering the specific issue. My real
curiosity is still why the performance I see regressed so badly when I
upgraded to F8. The thread leads me to believe that the the F7 swsusp1
did not save caches. Hmm... I have been enjoying the convenience of
fedora's suspend-works-out-of-the-box for awhile now, but I think it
might be time for me to go back to tux-on-ice. I truly am disgusted by
having to suffer through 5-15 seconds of thrashing while changing
desktops after resume. I would much rather the resume take 20 seconds
longer, and present me with a good user experience. (yes, I am one of
those people that thinks that offering early login while the system
finishes booting is a really stupid thing).


-dmc

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 07:03 AM
Douglas McClendon
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Douglas McClendon wrote:

[ pro-tux-on-ice rant snipped ]

might be time for me to go back to tux-on-ice. I truly am disgusted by
having to suffer through 5-15 seconds of thrashing while changing
desktops after resume. I would much rather the resume take 20 seconds
longer, and present me with a good user experience. (yes, I am one of
those people that thinks that offering early login while the system
finishes booting is a really stupid thing).


To clarify this harsh, theoretically offensive statement- Early login
done _right_ would be fine. Doing it the incomplete way, such as the
vista experience, is IMO a regression, not a feature. I.e. it boggles
my mind that fast boot-to-login time is perceived as so valuable, that
the implementers will let a user's first experience with the system be
at its absolute valley(anti-peak) responsiveness phase.


I guess for early login this isn't so bad, as people just learn that the
fast boot time is an illusion and that it's better to sit back and let
the system IO settle before using it. But what started my rant was the
swsusp choice, which is not workaroundable by just letting the system IO
settle.


and the 5-15 seconds was a bit of an exageration, more like 5-10. But
still, evocative of winblowz 3.1 swap thrashing....


</rant>

-dmc

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 04:45 PM
nodata
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Am Donnerstag, den 03.01.2008, 02:03 -0600 schrieb Douglas McClendon:
> Douglas McClendon wrote:
>
> [ pro-tux-on-ice rant snipped ]
>
> > might be time for me to go back to tux-on-ice. I truly am disgusted by
> > having to suffer through 5-15 seconds of thrashing while changing
> > desktops after resume. I would much rather the resume take 20 seconds
> > longer, and present me with a good user experience. (yes, I am one of
> > those people that thinks that offering early login while the system
> > finishes booting is a really stupid thing).
>
> To clarify this harsh, theoretically offensive statement- Early login
> done _right_ would be fine. Doing it the incomplete way, such as the
> vista experience,

I don't know why people keep saying this. Vista boots fast and is
responsive.

> is IMO a regression, not a feature. I.e. it boggles
> my mind that fast boot-to-login time is perceived as so valuable, that
> the implementers will let a user's first experience with the system be
> at its absolute valley(anti-peak) responsiveness phase.
>
> I guess for early login this isn't so bad, as people just learn that the
> fast boot time is an illusion and that it's better to sit back and let
> the system IO settle before using it. But what started my rant was the
> swsusp choice, which is not workaroundable by just letting the system IO
> settle.
>
> and the 5-15 seconds was a bit of an exageration, more like 5-10. But
> still, evocative of winblowz 3.1 swap thrashing....
>
> </rant>
>
> -dmc
>

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 05:23 PM
Jesse Keating
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

On Thu, 03 Jan 2008 18:45:16 +0100
nodata <lsof@nodata.co.uk> wrote:

> I don't know why people keep saying this. Vista boots fast and is
> responsive.

Not on my brand new shiny Dell M1330. 2 gigs of ram, a fast core2duo
cpu, and Vista was very slow to boot up (win2k speeds...) and once
booted and I logged in it was another long wait for everything to
settle down. Even then it was pretty unresponsive.

--
Jesse Keating
Fedora -- All my bits are free, are yours?
--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 05:53 PM
nodata
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

Am Donnerstag, den 03.01.2008, 13:23 -0500 schrieb Jesse Keating:
> On Thu, 03 Jan 2008 18:45:16 +0100
> nodata <lsof@nodata.co.uk> wrote:
>
> > I don't know why people keep saying this. Vista boots fast and is
> > responsive.
>
> Not on my brand new shiny Dell M1330. 2 gigs of ram, a fast core2duo
> cpu, and Vista was very slow to boot up (win2k speeds...) and once
> booted and I logged in it was another long wait for everything to
> settle down. Even then it was pretty unresponsive.

Mine boots in about 45 seconds, and then I launch putty and get on with
some work. Are you running something heavyweight?

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 06:02 PM
Jesse Keating
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

On Thu, 03 Jan 2008 19:53:30 +0100
nodata <lsof@nodata.co.uk> wrote:

> Mine boots in about 45 seconds, and then I launch putty and get on
> with some work. Are you running something heavyweight?

I'm not running it at all. Dell could have stuffed it full of crap,
who knows. I don't use windows.

--
Jesse Keating
Fedora -- All my bits are free, are yours?
--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 
Old 01-03-2008, 09:57 PM
Douglas McClendon
 
Default f8 gripe#2: why did f8's pm-hibernate regress?

nodata wrote:

Am Donnerstag, den 03.01.2008, 13:23 -0500 schrieb Jesse Keating:

On Thu, 03 Jan 2008 18:45:16 +0100
nodata <lsof@nodata.co.uk> wrote:


I don't know why people keep saying this. Vista boots fast and is
responsive.


lol.


Not on my brand new shiny Dell M1330. 2 gigs of ram, a fast core2duo
cpu, and Vista was very slow to boot up (win2k speeds...) and once
booted and I logged in it was another long wait for everything to
settle down. Even then it was pretty unresponsive.


Mine boots in about 45 seconds, and then I launch putty and get on with
some work. Are you running something heavyweight?



To be completely fair to microsoft, I'll admit that my issues may also
be heavily influenced by sony's participation in the spin of vista that
came with my laptop.


Also, my point was more about *first impressions* the user has with an
OS. I imagine that once vista has had a day to churn on my laptop after
fresh install/factory-recover, that it will boot much faster, and be
more responsive more quickly.


But when a new user buys a laptop like mine, turns it on for the first
time, and logs in the first time, it quite seriously has HOURS of high
IO churn that it needs to get through. As a result, a users first
experience with the OS, is at its worst. Not a great design tradeoff
choice IMO.


Which really makes no sense, as a factory recover shouldn't need to
precompute anything or build any initial databases that are going to be
exactly the same on every other piece of identical hardware. But yet,
that seems to be what is happening...


-dmc

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list
 

Thread Tools




All times are GMT. The time now is 10:38 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org