I cant test this patch by doing tests that go into swap. You add a
variable that wasn't used in the 100MB tests and going into swap would
no doubt cause major performance drops from this 100MB test no matter
what VM you used. The added overhead is very dependent on too many
variables to compare results. It's best to stay in memory with the
benchmark. Possibly with many small files instead of one large one and
disable swap altogether. Most users and developers dont deal with
400MB files anyway, they deal with many many small files most not larger
than 5MB. a benchmark of dealing with these would be much more
meaningful than finding some extreme example to show a weakness that is
obviously going to be there because of physical restricitions alone.
Oh, and btw, i ran the test anyway and it was lower than when i had run
it with the normal swap partition on the same drive. So the fs overhead
is a huge factor .... the lesson of the day, use swap partitions ....not
swap files
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Sep 07 2000 - 21:00:16 EST