lkml.org 
[lkml]   [1998]   [May]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectSwapping thoughts
My current configuration has 64 MB Ram and 128 MB swap. The current
kernels run pretty much OK, it's just that the swapping behavior
leaves a lot to desire.

I recently had a situation where I had ~40 MB buffers and the system
neverheless used ~12 MB of swap. After a while, the swap usage grew
although there was 'free' memory available.

IIRC, Linus said earlier that the current kernels run better if
they have some swap space to use. Anyway, the current behavior presents
some problems. I did some CD burning at the time, so that's why
I had tons of memory in buffers. I also had X running with several
netscape windows open. The fun starts when I switch to a netscape
window when that particular instance is swapped out. The switch
is like a crawl.. Somehow reminds me of Solaris, which is definitely
not very good. This kind of situation probably could be avoided
if the CD burner could tell the kernel that the files don't need
to be buffered (a useful feature anyway). But in this case that would
cure only the consequence of the (IMHO bad) swap behavior, not the
problem itself.

I remember the early days when a friend of mine praised the advantage
of a dynamic buffer cache, saying that 'all so called free memory
acts as cache and whenever an application needs more memory, the
cache is shrinked.' I just would like to have that behavior back.
The unneeded swapping causes bad task switch behavior and also
possibly jeopardizes a CD burn or any other io-bound application
since there is more disk activity, some of which could be avoided.
Granted, shrinking the buffer cache might also cause more disk
activity, but I don't believe it is reasonable to put so much weight
on buffering versus swapping sleeping processes out, since it feels
ridiculous to swap 20 MB if there are 45 MB of buffers.

Just my 0.02.

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

\
 
 \ /
  Last update: 2005-03-22 13:42    [W:0.043 / U:2.412 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site