Subject: Re: Discussing the future of the NetBSD scheduler
To: Martin Husemann <martin@duskware.de>
From: Eduardo Horvath <eeh@NetBSD.org>
List: tech-kern
Date: 03/09/2005 17:14:31
I'd ask what you think needs to be changed in the current scheduler
first. Although the current scheduler could be enhanced, it's
nicer than some alternatives.
On Wed, Mar 09, 2005 at 10:05:53AM +0100, Martin Husemann wrote:
> On Wed, Mar 09, 2005 at 12:27:52AM +0100, Daniel Sieger wrote:
> > or something more
> > sophisticated like a pluggable scheduler framework found in Solaris or
> > Linux?
>
> I'm with Bill that we probably want something like that in the longer
> time frame, but that the first step is to minimize the MD parts.
One problem with the Solaris scheduler is that the timeshare
scheduler instead of being a decaying priority scheduler turned
out to be more of a random priority scheduler. The nice thing
about the BSD scheduler is that it does get the decaying priority
bit right most of the time although it is a bit expensive. I don't
know if you can do all that math in a plugin.
Most of the neat features that could be added would be in the
real-time or SMP areas. But that will require kernel preemption
and fine-grained locking which would result in many years of
teething problems like Solaris and FreeBSD.
Eduardo