Re: 2.2.0-final

Nimrod Zimerman (zimerman@deskmail.com)
Sun, 24 Jan 1999 12:47:24 +0200


On Wed, Jan 20, 1999 at 11:10:42PM -0800, Linus Torvalds wrote:

Hello.

I was hoping that this issue resolves itself before 2.2.0 (with Andrea
Arcangeli's work), but it has not.

Problem: On small memory systems, the cache is too aggressive, swapping out
process memory, exchanging it with a lot of cache memory, for no good reason.
This causes obvious problems of processes being swapped out from beneath the
user (needlessly), which produces extremely bad performance and
interactivity compared to previous kernels (before pagecache and buffermem
limits were maimed).

A short term solution would be to put pagecache and buffermem limits back,
at least for those users who wish to have them (and maybe enable them by
default on small memory machines, as the defaults were before). I have no
idea how complicated this might be with the MM changes, though.
A long term solution would be to fix this, somehow, which is what Andrea
Arcangeli has been trying to do recently (his work improved things quite a
bit, but it is still worse than before).

Is it a good idea to 'release' 2.2.0 with such a problem? Linux is generally
known for being usable on low-end machines. Right now - it lacks in that
respect.

> "Is this something that normal people would ever really care deeply
> about?"

Probably, if there are people other than me that are still using systems
with 16mb of physical RAM. There probably are some. Maybe.

Nimrod

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