On Wed, 14 Jun 2000, Rik van Riel wrote:
>On Tue, 13 Jun 2000, Paul Barton-Davis wrote:
>
>> If Rik or Juan can assure me that streaming I/O performance is a
>> high priority for them, I might feel better, but I see more
>> concern about boundary cases caused by quite different
>> application patterns.
>
>Before I start hacking on this, could you tell me if 2.4.0-ac17+
>gives you the performance you want?
>
>(or at least, if it reaches 80% of 2.3.51 performance, the rest
^^^^^^
>seems related to the new elevator code and not to MM)
In general if you have any performance problem with the elevator:
1) use elvtune -b 128
2) increase read and write latency to big number like 100000
and let me and Jens know.
However IMHO the I/O performance issues are not related to the elevator.
There may be a performance drop due the I/O scheduler (but that's the
latency fix and we wanted it also in 2.2.x) but that shouldn't be a major
performance issue and for sure it can't change since 2.3.99-pre1. Latest
changes in the elevator happened in 2.3.99-pre1. And 2.3.51 had a much
worse elevator than what we have in the latest kernels. 2.3.51 was just
with the I/O scheduler included (but as said a for sure worse one).
So I'd appreciate if you could also explain which are the reasons that
makes you to think the elevator is the cause of the bad I/O performance
(and not the latest VM stuff).
Andrea
PS. I also have numbers using reiserf+classzone on ac7 and they are fine
as far I can tell (and nothing in the elevator is been changed for such
bench as far I can tell).
-
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/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:33 EST