Re: Exaggerated swap usage

Rik van Riel (riel@conectiva.com.br)
Fri, 29 Nov 2002 23:55:24 -0200 (BRST)


On Sat, 30 Nov 2002, Javier Marcet wrote:

> >First, lets get one thing straight: the problem is the slowness,
> >not necessarily the swap usage. It is very easy to jump to wrong
>
> OK, you might be right on this point.

> root # vmstat 1
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 0 1 265048 5248 32248 119108 6 15 65 62 252 585 21 6 73 0
> 1 6 266648 4480 32316 120300 0 4656 2152 4652 1348 821 13 8 79 0
> 1 0 265052 4496 31036 120184 8 336 1668 340 1226 765 15 7 78 0
> 0 1 265052 4496 31112 121564 4 0 3152 0 1198 894 18 8 74 0
> 1 0 265052 4504 31076 123112 0 0 3024 8576 1229 857 17 7 76 0
^^^^^^^^^^^^^^^^^^^

Looks like my guess was right after all. The amount of swap
IO is maybe 10% of the amount of filesystem IO in your vmstat
snippet above.

> This is after the system has been in use with a 512MB swap partition for
> around 1 hour. I must say it is barely usable as a desktop, way _far_
> from a responsive environment. looking at the memory numbers it's easy
> to think I need more memory, but with other kernels,

> So yes, you are right that swap usage is not the problem. It seems more
> like memory getting too dirty.

Two things could be happening here:

1) the kernel decides to cache the wrong things in the
page cache

and/or

2) the IO scheduler is giving worse latencies now

If the problem is (1) it might get resolved by using the -rmap
or -aa kernels. If the problem is (2) you'll want Andrew Morton's
read_latency patch (which I'll port to 2.4.20 real soon now).

regards,

Rik

-- 
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/		http://guru.conectiva.com/
Current spamtrap:  <a href=mailto:"october@surriel.com">october@surriel.com</a>

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