Re: [ck] swap prefetching merge plans

From: Grzegorz Kulewski
Date: Thu Mar 23 2006 - 02:53:45 EST


On Thu, 23 Mar 2006, Con Kolivas wrote:
On Thu, 23 Mar 2006 03:53 pm, Andrew Morton wrote:
A look at the -mm lineup for 2.6.17:

mm-implement-swap-prefetching.patch
mm-implement-swap-prefetching-fix.patch
mm-implement-swap-prefetching-tweaks.patch

Still don't have a compelling argument for this, IMO.

For those users who feel they do have a compelling argument for it, please
speak now or I'll end up maintaining this in -ck only forever.

Well, it works for me rather good. After I close some memory demanding aplication (like game or qemu) that caused huge portion of the memory to be swapped out to disk, the time needed to swith back to firefox or eclipse is really really shorter. Also the GUI is more responsive (because there are nearly none disk accesses needed when I am entering some dialog box or menu).

Without the patch system is taking up to several seconds to come back and even after that is feels jerky because it still must get some page from disk while I do some work.

Also I would really be glad to see the second part of the idea: mirroring some pages from memory to disk while system is relatively idle. This could make swapout nearly a no-op and this could really help some workloads if you are short on RAM. And software engeneerers and programmers are really working hard to eat all your RAM even if you think that "now you have really much".

I have Athlon XP 3000+ system with 1GB of RAM and 4GB of swap. It works pretty well on that configuration. I have so much swap because I am compiling I my packages (Gentoo) on tmpfs. So during (or after) some big compilation (like qt, gcc or something) swap prefetching is really helpful in getting my desktop apps back. (Also Con's scheduler is helpfull in allowing me to compile things in the background without nearly any problems and bad interaction with eclipse, openoffice and firefox.
It does not work excelent in all cases but it usually works very well.)

So, not that I have any vote here, I would really glad to have it merged. More people will be able to read the code and maybe it will become even better.


Thanks for your work,

Grzegorz Kulewski

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/