Re: Single task swap performance

Doug Paul (dbp@dragonsys.com)
Wed, 17 Apr 1996 17:27:40 -0400


Linus wrote
>On 15 Apr 1996, really kuznet@ms2.inr.ac.ru wrote:
>> Linux really has POOR swapping code, and kswap improved
>> it insignificantly. This opinion is result of every day experience
>> rather than benchmarks. I have Pentium 133 64Mb AIC7881 and run program
>> that eats >=128Mb of virtual memory (certainly, matrices).
>> It sweaps through these matrices with period ~20min.
>
>Agreed. This is one case where linux doesn't do anything clever at all,
>and one place where we should probably do some work.
[...]

Would also like to chime in...
I keep a Linux 1.3.60 128MB P5-90 (P6-200 on order) running round-
the-clock research jobs up to 450MB in size. The larger ones page
_very_ poorly. In some cases (linear pass through a large array),
I only get 2% user cpu time. I also do my interactive work on the
same machine.

Kswap appears to have helped interactive performance during swapping,
but I have no data on the effect on the large job itself.

I too would like to second the suggestion that some form of
improved swapping for single large jobs be considered in the
not-too-distant future.

Certainly appreciate the effort so far--thanks to all who have
contributed.

Doug