BFS Scheduler Benchmarks

[kraftman]

I found it interesting that BFS outperformed CFS on serving Apache pages, that one is sure to interest some people. Especially if it holds up with more processors.

CFS is faster with fair_sleepers disabled. At least here.

[discordian]

@Discordian

UI responsiveness is a different thing. It's sad, but even Win95 had very good 2D acceleration. When comes to audio skipping I have much better experience with Linux then with Windows (it seems improperly configured Pulse Audio can introduce some problems, but I don't use it)

No, its related. Without low latency scheduling there is no responsive UI, raw power is also needed, but not the issue as this ain't notable better with faster hardware.
And I went through pains disabling Pulse Audio some time ago, its not like I haven't tried for years. If nothing else it made me appreciate Windows

Anyway, if you don't expect UI-responsiveness and Media-Playback to benefit from low latency.. then what should benefit in your opinion?

[mutlu_inek]

@Mutlu_inek

It will be good to see some some benchmarks which measure responsiveness, but what benchmark will show you this? However, maybe there are such tests. For some people BFS is more responsive and for some others it's not. It will be great to have some problems fixed in both schedulers.

Not having a variety of benchmarks measuring responsiveness is the major problem here. I do not expect Phoronix or anyone else here to develop a new test just for BFS. I don't even know whether BFS is actually better than CFS. But if I was to benchmark BFS, I would put a huge disclaimer at the beginning and the end of the test in order to put in perspective that the benchmarks don't actually address the issue at all.

Con Kolivas is, of course, pointing out that we s/could use Interbench. He wrote the test himself (before creating BFS) and also uses it to test his work. The benchmark is available for most distributions, AFAIK.

[cnyk]

(...)
Benchmarks measuring performance are detrimental to Con's effort as they reinforce the way things have been done before: get the last bit out of the cpu, at the expense of the user experience. Being fast in absolute terms is making my experience of the system worse as I am annoyed by unresponsive UIs or dropped frames while I am unable to measure whether a certain process took a few more seconds to complete. (...)

I completely agree. It's quite irrelevant how long it takes for, let's say, VLC to compile (that's what my system is doing while I'm writing this). It is however very important to me (and to the purpose of BFS) what the following "test" shows:

Yesterday after reading the Phoronix post about BFS, and whatever Con had to say about his latest piece of code, I started compiling the kernel to see what BFS is good for. In the meantime I was trying to watch Federer's wonderball on youtube. I had to wait for the kernel compilation to finish.

Now I'm watching the video in Opera, writing this in FF and my system is compiling VLC in the same time. FF scroll action is flawless, the video playback is flawless, vlc is happily compiling in the background. The only hickup is task switching, but I think that's KDE's fault not the BFS.

So I guess that no matter what that speed tests say, for me at least BFS has reached it's target. Many thanks to Con Kolivas for doing it, and to Michael for telling me (us ) about it.

[kraftman]

No, its related. Without low latency scheduling there is no responsive UI, raw power is also needed, but not the issue as this ain't notable better with faster hardware.
And I went through pains disabling Pulse Audio some time ago, its not like I haven't tried for years. If nothing else it made me appreciate Windows

Anyway, if you don't expect UI-responsiveness and Media-Playback to benefit from low latency.. then what should benefit in your opinion?

If you had nice accelerated 2D under Linux you probably wouldn't notice difference in UI-responsiveness between BFS/CFS/CFS_with_sleepers_disabled . However, it's good to tune schedulers while it's still easy to notice some differences or problems (mainly thanks to slow 2D ). Of course, compilation time and some other things are not 2D related.

Programs like lmms should benefit. AFAIK low latency != visible latency and I wonder if BFS is made for low latency like RT kernels.

@Mutlu_usek

Not having a variety of benchmarks measuring responsiveness is the major problem here. I do not expect Phoronix or anyone else here to develop a new test just for BFS. I don't even know whether BFS is actually better than CFS. But if I was to benchmark BFS, I would put a huge disclaimer at the beginning and the end of the test in order to put in perspective that the benchmarks don't actually address the issue at all.

Yes, this will be sane.

[kUrb1a]

I agree with all of this.One thing that I was interested when Ubuntu 7.10 was released is how to bring my old PC back to life.He was struggling with every day work and something had to be done.

I always wanted to try out linux and everybody back then who know about linux was telling me how fast and reliably it is.

Now i know it's not true at all.Yes if you don't wan't anything on screen except command line but that's was not the case with me.

Back then i had ATI card and it was a bit** to make it work as it is now but when I did overall usefulness of Ubuntu was not improved because it was slowwwwww,...like computer was struggling and that bum d me down because linux supposed to be everything but that and sad part it's still the problem it's not smooth at all.

Anyway I would love to test out this patch and run some test's to see how it's stacking up with CFS.I was wondering if anybody here know what test's should I run i have 2 computers it can run as long it takes.

[kraftman]

Now i know it's not true at all.Yes if you don't wan't anything on screen except command line but that's was not the case with me.

So, 2D is slow for you?

Back then i had ATI card and it was a bit** to make it work as it is now but when I did overall usefulness of Ubuntu was not improved because it was slowwwwww,...like computer was struggling and that bum d me down because linux supposed to be everything but that and sad part it's still the problem it's not smooth at all.

Have you ever tried some other distro? Ubuntu has some scripts which can make it slow. Btw. what exactly was slow for you? If its interface it's nothing new

Anyway I would love to test out this patch and run some test's to see how it's stacking up with CFS.I was wondering if anybody here know what test's should I run i have 2 computers it can run as long it takes.

Maybe you could try some random Phoronix benchmarks? :> If you don't mind can you try running some 3D game (some fps like Urban Terror should be great) and listen to the music same time using Amarok or Rhythmbox and tell if game slow downs with the BFS scheduler?


Probably the most interesting will be just your personal feeling. If BFS will seems to be better can you also try this with CFS (as root):

mkdir /debug
mount -t debugfs none /debug
echo NO_NEW_FAIR_SLEEPERS > /debug/sched_features

it helps on some configurations when comes to responsiveness and this change will also have influence on some benchmarks results.

[kUrb1a]

Well pretty much 2D especially when it comes to fglrx somehow it always breaks something or makes it right but slow.Any way that was years earlier now that OSS team does such great work I can't really complain.

Slow 2D I can forget but it's on my so-could new computer with q6600 and 4Gb of ram.Things are not so bad when I use it lightly on computer like when I study for college,just open office,netbeans and maybe songbird but god forbid that I open more than 15 tabs in firefox or opera.I tried all available patches,how-tos.

And kraftman I have installed on my computer all bunch of games including like HoN,UrbanT,Nexuiz and yes I tried like 12 distros so far and I always take a look at new releases(not virtualization )and I always get disappointed.So far Arch feels very appealing .I will try as many CPU bound test I can maybe,BFS is my solution .I'll try to post tomorrow results.

BTW,patched 2.6.30,6 kernel with zen5 patch,BFS 220

[kraftman]

Well pretty much 2D especially when it comes to fglrx somehow it always breaks something or makes it right but slow.Any way that was years earlier now that OSS team does such great work I can't really complain.

Slow 2D I can forget but it's on my so-could new computer with q6600 and 4Gb of ram.Things are not so bad when I use it lightly on computer like when I study for college,just open office,netbeans and maybe songbird but god forbid that I open more than 15 tabs in firefox or opera.I tried all available patches,how-tos.

Imagine 15 tabs with unaccelerated flash animations in each tab . However, everything should be better if something called Gallium3D will give us accelerated desktop (or at least when flash will be accelerated).

And kraftman I have installed on my computer all bunch of games including like HoN,UrbanT,Nexuiz and yes I tried like 12 distros so far and I always take a look at new releases(not virtualization )and I always get disappointed.So far Arch feels very appealing .I will try as many CPU bound test I can maybe,BFS is my solution .I'll try to post tomorrow results.

BTW,patched 2.6.30,6 kernel with zen5 patch,BFS 220

That's a lot of games hehe. Yes, Arch is very appealing

[aigarius]

I am amazed at the web performance results. Could that be expanded into more real-life situations: static pages with nginx, full LAMP stack (with PHP, Perl and Python), django, turbo gears, ruby on rails, Tornado, ...

In high end web business people spend millions just to get a few percent of performance gain, if BFS or a variant of it can give a 10% speed boost in heavily loaded web serving situation, it would be a godsend and would be pushed as an option into mainline very, very rapidly.