On Sun, 23 Apr 2000, Jens Axboe wrote:
>
> > Could this be the cause of the relatively bad results on hdrbench compared to
> > FreeBSD ?
> > (see my previous post about hdrbench)
>
> I don't know if FreeBSD can do TCQ on IDE, but the current elevator
> in 2.3 does not give very good performance results. I'm experimenting
> on improving this and results look very promising so far. Expect
> more soon.
Nice, I am looking forward for the results.
Notice that I tested kernel 2.2.12 (stock RH6.0) , 2.2.14 ( stock RH6.1) and
2.3.99-pre5, basically all give me the "wild" curves. (big peaks = long stalls
of the disk thread)
You can use hdrbench testings, since its simultanous reading/writing to many
files at the same time, generates quite a nice stress to the IO subsystem,
and looking at the graph you can quickly see if the elevator/request
reordering/ buffercache flushing are working well.
(ideally the curves should be as flat as possible, indicating that both the
requested transferrate can be achieved, plus the ongoing number of
read-writes/time remanis constant, without any long stalls (I measured up to
8secs or so).
Benno.
http://www.linuxdj.com/hdrbench
-
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 : Sun Apr 23 2000 - 21:00:22 EST