Subject: Re: Throttling IO Requests in NetBSD via Congestion Control
To: Jukka Marin <email@example.com>
From: Bill Studenmund <firstname.lastname@example.org>
Date: 08/21/2006 13:59:11
Content-Type: text/plain; charset=us-ascii
On Mon, Aug 21, 2006 at 11:52:57PM +0300, Jukka Marin wrote:
> On Mon, Aug 21, 2006 at 01:35:03PM -0700, Bill Studenmund wrote:
> > As one of the mentors for this project, I think changing scheduling
> > priority can/should be investigated after the SoC project completes.
> If we have two (or more) I/O bound processes running, CPU usage is very
> low and the disk is 100% busy - I think changing the scheduling priorities
> would have little or no impact on the system performance, right? There's
> no shortage of CPU cycles, so no matter how low the priority of any of the
> processes is, the process will always get the few cycles it needs to queue
> yet another I/O request to the disk system.
> Not that putting a process to sleep sounds that much better.. :-)
The scenario you describe isn't the one that this work was targeting. This=
work is aimed at a case where there is/are reader(s) along with the=20
writer(s). So the work is to skew disk access towards them.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (NetBSD)
-----END PGP SIGNATURE-----