On Wed, 31 May 2000, Michal Ostrowski wrote:
> I've noticed some horrible I/O performance in recent 2.3
> kernels. My first guess was that this was related to the
> various VM problems that have been running rampant recently, but
> now I'm not so sure. Even though I've been reading reports that
> VM performance has been improving, I've seen no noticeable
> impact on my test results.
In fact, 2.3.99-pre9 doesn't contain most of the "new VM" stuff,
that went in in the -ac* series (and seems to have increased
performance very slightly).
(most test results snipped for brevity)
> Celeron 500 Dual PIII 550
> test1-ac7 2.3.99-pre9
> Threads Blocks Time To Complete 1000 Reads (seconds)
> 1 32 22.0 32.3
> 4 32 20.2 28.5
> 10 32 290 345 *
The fact that performance really deteriorates when you
run more threads suggests that this may have something
to do with the elevator code.
> * 2.2.14 runs this test in 34 seconds.
How fast are 2.2.15 and the latest 2.2.16pre kernel?
The elevator code changed after 2.2.14, so it would
be an ideal testbed for seeing what the culprit is.
(VM changed too, but in a completely different way
from how 2.3/2.4 VM changed)
-- The Internet is not a network of computers. It is a network of people. That is its real strength.
Wanna talk about the kernel? irc.openprojects.net / #kernelnewbies http://www.conectiva.com/ http://www.surriel.com/
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to firstname.lastname@example.org Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:12 EST