state of Fedora kernel and Desktop use?

[Valent Turkovic]

Hi devels & all,
I'll be brief so not to take much of your time.

I read this article [1] about general Linux kernel development and how
desktop use is ignored by mainstream kernel developers, and how
desktop performance is suffering because of that.

I took out this quote to summarize:
"I think the kernel developers at large haven't got the faintest idea
just how big the problems in userspace are. It is a very small brave
minority that are happy to post to lkml, and I keep getting users
telling me on IRC, in person, and via my own mailing list, what their
problems are. And they've even become fearful of me, even though I've
never viewed myself as a real kernel developer."

What is the status of Fedora kernel regarding specific patches that
make desktop perform better? Is is true that mostly big vendor issues
get resolved and that kernel devels don't care enough for desktop
users?

Cheers,
Valent.

[1] http://apcmag.com/interview_with_con_kolivas_part_1_computing_is_boring.htm

-- 
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic

[Chris Snook]

Valent Turkovic wrote:
>  Hi devels & all,
>  I'll be brief so not to take much of your time.
>  
>  I read this article [1] about general Linux kernel development and how
>  desktop use is ignored by mainstream kernel developers, and how
>  desktop performance is suffering because of that.
>  
>  I took out this quote to summarize:
>  "I think the kernel developers at large haven't got the faintest idea
>  just how big the problems in userspace are. It is a very small brave
>  minority that are happy to post to lkml, and I keep getting users
>  telling me on IRC, in person, and via my own mailing list, what their
>  problems are. And they've even become fearful of me, even though I've
>  never viewed myself as a real kernel developer."
>  
>  What is the status of Fedora kernel regarding specific patches that
>  make desktop perform better? Is is true that mostly big vendor issues
>  get resolved and that kernel devels don't care enough for desktop
>  users?
>  
>  Cheers,
>  Valent.
>  
>  [1] http://apcmag.com/interview_with_con_kolivas_part_1_computing_is_boring.htm
>   
We do plenty of desktop optimization.  Con Kolivas was whining because we were 
being conservative about desktop optimizations that had a high risk of harming 
enterprise workloads, but the truth is that we're conservative about all 
optimizations that have a high risk of harming *any* workload, which is how we 
should be.

As for specific patches, we try to keep as close in sync with upstream as 
possible.  Fedora-specific patches are more likely to be features that aren't 
quite ready for the vast range of kernel configurations possible upstream, but 
we're confident of in the limited configurations we use in Fedora.  Performance 
optimizations tend to track rather closely.  Some distributions use experimental 
patches that haven't been merged upstream to optimize for some specific 
workload, but Fedora is used far too widely for that, and the payoff of those 
patches is generally rather small anyway.  For things that make a huge 
difference, there's enough motivation to get them upstream quickly, so we get 
them quickly too.

-- Chris

[Josh Boyer]

On Fri, Oct 24, 2008 at 01:05:00PM +0200, Valent Turkovic wrote:
> Hi devels & all,
> I'll be brief so not to take much of your time.
>
> I read this article [1] about general Linux kernel development and how
> desktop use is ignored by mainstream kernel developers, and how
> desktop performance is suffering because of that.
>
> I took out this quote to summarize:
> "I think the kernel developers at large haven't got the faintest idea
> just how big the problems in userspace are. It is a very small brave
> minority that are happy to post to lkml, and I keep getting users
> telling me on IRC, in person, and via my own mailing list, what their
> problems are. And they've even become fearful of me, even though I've
> never viewed myself as a real kernel developer."
>
> What is the status of Fedora kernel regarding specific patches that
> make desktop perform better? Is is true that mostly big vendor issues 
The status is the same here as it is for every other kind of patch.
Fedora tries very hard to not carry additional patches.  With the
exception of a few things like exec-shield, if the patches aren't
upstream or headed there rapidly, they generally aren't carried.

It's easier to talk about specific patches when they are pointed out.
Neither the interview, nor you did that so I don't really know what
sort of patches you're talking about.

josh

[Valent Turkovic]

On Fri, Oct 24, 2008 at 1:59 PM, Josh Boyer <jwboyer(a)gmail.com&gt; wrote:
>  On Fri, Oct 24, 2008 at 01:05:00PM +0200, Valent Turkovic wrote:
> >Hi devels & all,
> >I'll be brief so not to take much of your time.
> >
> >I read this article [1] about general Linux kernel development and how
> >desktop use is ignored by mainstream kernel developers, and how
> >desktop performance is suffering because of that.
> >
> >I took out this quote to summarize:
> >"I think the kernel developers at large haven't got the faintest idea
> >just how big the problems in userspace are. It is a very small brave
> >minority that are happy to post to lkml, and I keep getting users
> >telling me on IRC, in person, and via my own mailing list, what their
> >problems are. And they've even become fearful of me, even though I've
> >never viewed myself as a real kernel developer."
> >
> >What is the status of Fedora kernel regarding specific patches that
> >make desktop perform better? Is is true that mostly big vendor issues
>
>  The status is the same here as it is for every other kind of patch.
>  Fedora tries very hard to not carry additional patches.  With the
>  exception of a few things like exec-shield, if the patches aren't
>  upstream or headed there rapidly, they generally aren't carried.
>
>  It's easier to talk about specific patches when they are pointed out.
>  Neither the interview, nor you did that so I don't really know what
>  sort of patches you're talking about. 
Interview focused on specific issues - "Staircase Deadline CPU scheduler"

to quote again from the artcle (page 2/3)

"The Staircase Deadline CPU scheduler. Initially started as a side
project from the Staircase CPU scheduler I soon realised that it was
possible to have excellent interactivity while fixing the horrible
fairness issues that an unfair design had. Furthermore, it actually
improved interactivity issues elsewhere that ended up being fairness
problems, and fairness is of course paramount to servers and multiuser
environments."

and also:

"One very vocal user found that the unfair behaviour in the mainline
scheduler was something he came to expect. A flamewar of sorts erupted
at the time, because to fix 100% of the problems with the CPU
scheduler we had to sacrifice interactivity on some workloads. It
wasn't a dramatic loss of interactivity, but it was definitely there.
Rather than use 'nice' to proportion CPU according to where the user
told the operating system it should be, the user believed it was the
kernel's responsibility to guess. As it turns out, it is the fact that
guessing means that no matter how hard and how smart you make the CPU
scheduler, it will get it wrong some of the time. The more it tries to
guess, the worse will be the corner cases of misbehaving."

Sorry for quoting, but I'm not a developer, just a curious user :)

Thank you very much for replying.

Valent.


-- 
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic

[Ignacio Vazquez-Abrams]

On Fri, 2008-10-24 at 14:31 +0200, Valent Turkovic wrote:
>  Sorry for quoting, but I'm not a developer, just a curious user
> :) 
And not to belittle your status as a user, but at some point you are
going to have to trust that the developers are Doing The Right Thing.

-- 
Ignacio Vazquez-Abrams <ivazqueznet(a)gmail.com&gt;

PLEASE don't CC me; I'm already subscribed

[Valent Turkovic]

2008/10/24 Ignacio Vazquez-Abrams <ivazqueznet(a)gmail.com&gt;:
>  On Fri, 2008-10-24 at 14:31 +0200, Valent Turkovic wrote:
> > Sorry for quoting, but I'm not a developer, just a curious user :)
>
>  And not to belittle your status as a user, but at some point you are
>  going to have to trust that the developers are Doing The Right Thing. 
I never said I don't trust them, far from it, just that I'm curious.

I'm really impressed how well Fedora 9 and current rawhide work, but
when I read CK's interview I was sad that he quit doing his patches.
Desktop linux (Fedora included) needs all the people that it can,
especially ones who focus also on the desktop and issues regarding
desktop use. He claimed that desktop performance was seriously
impaired when 2.6 kernel was released and that then nobody was even
paying attention on desktop workloads and that enterpise workload
patches were seriously hurting desktop performance and workloads.

If there is some other resource  (his personal blog, interview or
something) about Linux kernel and desktop workloads that that some
kernel developer wrote please point me in that direction.

Cheers,
Valent.

-- 
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic

[Jeff Moyer]

"Valent Turkovic" <valent.turkovic(a)gmail.com&gt; writes:

>  2008/10/24 Ignacio Vazquez-Abrams <ivazqueznet(a)gmail.com&gt;:
> > On Fri, 2008-10-24 at 14:31 +0200, Valent Turkovic wrote:
> >> Sorry for quoting, but I'm not a developer, just a curious user :)
> >
> > And not to belittle your status as a user, but at some point you are
> > going to have to trust that the developers are Doing The Right Thing.
>
>  I never said I don't trust them, far from it, just that I'm curious.
>
>  I'm really impressed how well Fedora 9 and current rawhide work, but
>  when I read CK's interview I was sad that he quit doing his patches.
>  Desktop linux (Fedora included) needs all the people that it can,
>  especially ones who focus also on the desktop and issues regarding
>  desktop use. He claimed that desktop performance was seriously
>  impaired when 2.6 kernel was released and that then nobody was even
>  paying attention on desktop workloads and that enterpise workload
>  patches were seriously hurting desktop performance and workloads. 
You do realize that this whole Linux thing started because some
developer wanted to run UNIX on his desktop, right?  That hasn't
changed.

http://www.techworld.com/opsys/news/index.cfm?newsid=9652

Cheers,

Jeff

[Linus Walleij]

fre 2008-10-24 klockan 13:05 +0200 skrev Valent Turkovic:

>  What is the status of Fedora kernel regarding specific patches that
>  make desktop perform better? 
The CFS was merged ages ago and there is a very wide consensus that it
gives the same kind of interactivity boost as CK:s rotating staircase.

On top of that the control groups for CPU slicing will be available in
F10. There is no consensus of how to make use for it in the desktop
scenarios though, so won't add much as for now. Control groups look like
they will be able to e.g. prioritize things like the memory hierarchy
and network traffic for the desktop use case (as well as other use
cases) in the long (well, some year or two) run, but it is being
intensely developed right now.

Swap prefetch was the most interesting patch from the CK tree IMHO:
http://lwn.net/Articles/242765/
Sadly I don't know what has happend in this regard :-/ Anyone else?

While we're at interactivity, there is this thing called "ionice" that
interacts with the CFQ I/O scheduler from userspace. It is probably a
good idea  to "ionice -c1" (realtime) things like your Gstreamer or
mplayer, and "ionice -c3" (idle) your transmission bittorrent client,
updatedb or backup tasks. I do this manually from time to time, but
cannot really claim I notice any big differences. Such heuristic
ionice:ing is nowhere to be seen, does anyone know if it is being looked
into anywhere at all?

All in all, there are popping up some real nice interactivity features
in the kernel, but AFAIK noone is really working on using them in the
desktop usecase.

Just my Euro 0.01...
Linus

[Bruno Wolff III]

On Fri, Oct 24, 2008 at 22:48:59 +0200,
  Linus Walleij <triad(a)df.lth.se&gt; wrote:
>  fre 2008-10-24 klockan 13:05 +0200 skrev Valent Turkovic:
>  
>  All in all, there are popping up some real nice interactivity features
>  in the kernel, but AFAIK noone is really working on using them in the
>  desktop usecase. 
It's not working so nice in rawhide. I am seeing copying data to an encrypted
usb drive lock up the X server for times on the order of a minute (presumably
when it needs to page something in). Very annoying. At least it recovers.
When combined with high memory use, sometimes the whole machine locks up
and a reboot is necessary to clear things up.