Re: 2.1.91 swap performance: jerky.

Linus Torvalds (torvalds@transmeta.com)
Fri, 27 Mar 1998 18:16:22 -0800 (PST)


On Sat, 28 Mar 1998, Itai Nahshon wrote:
>
> The idea sounds right, I have questions regarding limitation on
> current implementation:
> 1. If something is swapped out but the pages that it occupied
> never used later, will it have to be paged-in from disk or can
> the copy in memory be reused without any farter io?

If the page is never used again, it will never be paged in: an exit of the
process will just free up the page on disk, and a "fork()" will duplicate
the pointer to the disk rather than page it in in order to duplicate it.

> 2. Do the background swapper compete with interactive processes
> on the disk bandwith?

Yes. This is obviously a concern, but hopefully the background swapper is
never _too_ aggressive, ie it shouldn't be aggressive enough for this to
be a real issue.

The parameters may need more tweaking, and there may be some other minor
changes in the air, but on the whole the basics should be there now.

Linus

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