[SOLVED] Famous "4 lines better than 200" patch

dcbdbis · 2011-06-06 04:20:36

I'm probably searching incorrectly. I can't find the answers to the below. I apologize if my answers are right in my face and I missed it.

I have a small netbook that I use as my test bed. Mint has a debian edition with this patch. When I installed this distro to my little netbook, it really, really perked it up.

Arch is my distro. I use the mainstream kernels. I do exclude new kernels in pacman.conf until they've been through a few patches....typicalically an ending number > .4.

I do occasionally put a tremendous I/O load on my Arch desktop in my backup process.....and in so doing my desktop is very slow to respond. Same thing when I allow VMWare to do a "Defrag" or "Compact" on my VM's. These processes put a heavy load on disk I/O.

I've looked at the AUR, and other than some references to "BF" type schedulers and the like....I don't see anything in the forums or in the AUR that jumps out at me.

So may I ask please:

a) Does Arch use this infamous patch in their mainstream kernels?

b) I've heard and read in various forums about different ways to accomplish the same desktop perkiness without actually doing a kernel changeout. Can someone please point me to this post/wiki/other info if it's really worthwhile (or the posts I read were legit)?

c) Does this infamously patched kernel exist in the AUR? If so, can someone please tell me the name?

d) If a reader of this post is of the opinion I should avoid kernels in the AUR for whatever reasons, I'd also like to hear that......and for my own knowledge, their reasoning.

It's not a big deal for me, as I don't do these massive movements of data around but once a week...... And I certainly do not want to destabilize my system....

I've built my own custom kernels before on Arch, Debian, and a handful of other distros. Aud-nauseum. I'd rather spend my time on something else than tweaking a .config file to gain a few ms....So I'm admittedly 'lazy' and use pre-config'd kernels by others, in my case, I use the mainstream kernel.

For the record, I always add to the fstab the "noatime" option for my mounted ext4 partitions. And in sysctl.conf I put the vm.swappiness=0 and the vm.vfs_cache_pressure=50 statements. I'm on x86_64 w/8Gb of ram and an Intel Quad extreme proc (just before the i7's hit the market).

A goal would be to have my desktop be more responsive under heavy kernel I/O load.

I do want to be clear though, It's a "want" not a "need".

Sincerely and respectfully,

Dave.

Last edited by dcbdbis (2011-06-08 00:44:02)

graysky · 2011-06-06 04:35:11

I think the patch to which you refer has been added to mainline in 2.6.38 hasn't it? If I am right then:

a) yes
b) no idea
c) no idea
d) I would recommend kernel26-ck (bfs) for optimal desktop responsiveness -- particularly on a C2Q! I run an X3360 (Xeon quad) and use kernel26-ck.

bangkok_manouel · 2011-06-06 04:35:11

eb64@drama:~$pacman -Q kernel26
kernel26 2.6.38.7-1
eb64@drama:~$zgrep -i autogroup /proc/config.gz
CONFIG_SCHED_AUTOGROUP=y

/edit: for b), see http://www.webupd8.org/2010/11/alternat … patch.html

Last edited by bangkok_manouel (2011-06-06 04:37:56)

brebs · 2011-06-06 06:20:50

I suggest reading:

https://bbs.archlinux.org/viewtopic.php?id=116481
http://forums.gentoo.org/viewtopic-t-869309.html

ChoK · 2011-06-06 11:43:47

You are confusing two issues.
The BFS and CFS schedulers schedules CPU loads. On heavy I/O loads, the processes are actually waiting for I/O to complete, renicing or reordering them won't helpthat much. What is weird is that a process waiting for I/O slows the whole desktop.

For information there is also I/O schedulers : BFQ, CFQ, noop and deadline. You can switch between them at boot time with the "elevator=" flag.

dcbdbis · 2011-06-07 23:29:37

Thank you for all the replies and the information. I now have the information I needed in order to make an informed choice.

I read the links provided, and additional links I located once I knew more about what I had originally posted.

I must say, that I'm disappointed that there appears to be some sort of pissing contest between the kernel devs and C.K. In my experience with different schedulers, they fit different purposes, and BFS appears to fill an additional scheduler need that's not available in mainstream at this point.

Linux is all about choice after all!

Again thank you for the information and the links. They are appreciated.

Dave..

dcbdbis · 2011-06-07 23:44:38

For Graysky:

I see in your pre-compiled repo a core2 optimized version.

Question: The original Core2 had only two cores, it was later that Intel added a Quad core version of the Core2. Forgive me for the next question, but I need to ask it:

May I assume that your Core2 optimized packages include proper support for Intel's Quad core version of the Core2?

And as I by necessity need to use VMWare Player over VirtualBox (for now, DirectX issues), do you foresee any issues rebuilding the VMWare kernel modules against your -ck-core2 kernel?

Thanks!

Dave...........

Last edited by dcbdbis (2011-06-07 23:47:02)

graysky · 2011-06-08 00:34:24

@dcd - correct, "core2" is the optimized build for any CPU in that "core2" family including quads (I run this kernel on mine). I haven't used vmware myself, but I assume that it has a script for building the needed modules as virtual box does which is independent of host kernel.

dcbdbis · 2011-06-08 00:43:36

It does indeed have it's own build scripts and I'm totally comfy with using them from the cli.

Thank you very much for the reply!

Dave............

rwd · 2011-06-08 21:06:11

As far as the backup causing lots of io is concerned : what kind of backup application/method do you use? rdiff-backup (incremental backups with backward deltas) with a low 'nice' value takes around 15 to 40 minutes here to backup to a nas but doesn't cause that much of a slow-down.

dcbdbis · 2011-06-09 02:18:23

rwd wrote:

...what kind of backup application/method do you use?

rsync.

1.82 TB's worth...........

Dave...

Last edited by dcbdbis (2011-06-09 02:18:56)

Arch Linux

#1 2011-06-06 04:20:36

[SOLVED] Famous "4 lines better than 200" patch

#2 2011-06-06 04:35:11

Re: [SOLVED] Famous "4 lines better than 200" patch

#3 2011-06-06 04:35:11

Re: [SOLVED] Famous "4 lines better than 200" patch

#4 2011-06-06 06:20:50

Re: [SOLVED] Famous "4 lines better than 200" patch

#5 2011-06-06 11:43:47

Re: [SOLVED] Famous "4 lines better than 200" patch

#6 2011-06-07 23:29:37

Re: [SOLVED] Famous "4 lines better than 200" patch

#7 2011-06-07 23:44:38

Re: [SOLVED] Famous "4 lines better than 200" patch

#8 2011-06-08 00:34:24

Re: [SOLVED] Famous "4 lines better than 200" patch

#9 2011-06-08 00:43:36

Re: [SOLVED] Famous "4 lines better than 200" patch

#10 2011-06-08 21:06:11

Re: [SOLVED] Famous "4 lines better than 200" patch

#11 2011-06-09 02:18:23

Re: [SOLVED] Famous "4 lines better than 200" patch

Board footer