Re: Kswapd future (was: Re: new kmod.c - debuggers and testers needed)

Rik van Riel (H.H.vanRiel@phys.uu.nl)
Wed, 15 Apr 1998 11:51:02 +0200 (MET DST)


On Tue, 14 Apr 1998, Perry Harrington wrote:

> Is there some way you could tap into the "idle" task processing time to do this?

Yes there is... We could reprogram the idle task to do some
special task instead of idling.
This shouldn't be to difficult...

> Solaris has a kernel thread which implements page scanning. It's CPU usage
> and aggressiveness is dictated by the slowscan, fastscan, maxpgio, and
> lotsfree.

OK, so we need to tune kswapd better, better and better...
We're working on it (well, I know I am).

> What is actually more intensive, doing page scanning and doing speculative
> paging, or doing desperation swapping on the fly?

Currently, my machine is doing both. I've added a high/low
water mark to kswapd's free_memory_available(), and the disk
is very idle now...
I've posted the patch to linux-mm and mailed it to Linus.
I'll put it on my homepage as well (RSN).

Rik.
+-------------------------------------------+--------------------------+
| Linux: - LinuxHQ MM-patches page | Scouting webmaster |
| - kswapd ask-him & complain-to guy | Vries cubscout leader |
| http://www.fys.ruu.nl/~riel/ | <H.H.vanRiel@fys.ruu.nl> |
+-------------------------------------------+--------------------------+

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu