On Fri, 4 Jun 1999 15:35:36 +0200 (CEST), Andrea Arcangeli
<andrea@suse.de> said:
> Here it is my idea that I am testing since some day ago. It completly
> avoid the DoS doing `cp /dev/zero /tmp'. Now a `find` go fast near as
> there wouldn't be a `cp /dev/zero /tmp' in background. It improves (or
> better it gives you :) the iteractive response while kflushd is doing its
> work. It seems to have no impact on normal write performances. It seems to
> only avoid the readers-starvation and I am quite happy with it.
Unfortunately it does completely the wrong thing if you have more than
one disk: it stalls all writes on all disks as soon as any one disk goes
into wait-on-read. That's not good. Not good at all.
This is precisely the problem with optimising for one micro-benchmark.
We really do need per-block-device IO scheduling, in some form or
another, to fix this once and for all. We already have the per-device
CURRENT queues, so some of the necessary infrastructure is in place
today (and the current mechanism does support devices specifying their
own queues to let them share queues sensibly).
--Stephen
-
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/