Re: VM tests on 2.4.15-pre6 and 2.4.15pre6aa1

[Andrea Arcangeli]

On Sun, Nov 18, 2001 at 09:57:20AM -0500, [e-mail] wrote:
> Hope this information is useful!

yes it's useful. It seems normal, as said this kernel isn't going to
swap very easily, it tends to shrink the cache too much, and I only
cared to fix the >=3G boxes with no swap at all but with lots of cache
where I wanted to be sure the anon pages into the lru weren't going to
make the system sluggish (google setup incidentally). I increased
significantly the vm_mapped_ratio default setting to be sure no to have
swap_out in our way during the testing (of course, increasing
vm_mapped_ratio doesn't mean the swap_out doesn't work, but only makes
harder to the machine to run into swap, it tends to shrink the cache
more).

So if you've time to test, I'd like to know what you get also after a:

	echo 9 >/proc/sys/vm/vm_mapped_ratio

and also after backing out those two patches in order:

	ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.15pre6aa1/10_vm-14-no-anonlru-1-simple-cache-1
	ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.15pre6aa1/10_vm-14-no-anonlru-1

and again using "echo 9 >/proc/sys/vm/vm_mapped_ratio" after backing out
the two patches.

thanks for the feedback,

Andrea