Re: [Lse-tech] Re: [PATCH][RFC] Proposal For A More Scalable Scheduler ...

[Mike K. <kr...@us...>]

On Tue, Oct 30, 2001 at 11:52:57AM -0500, Hubertus Franke wrote:
> * Davide Libenzi <da...@xm...> [20011030 13;50]:"
> > On Tue, 30 Oct 2001, Hubertus Franke wrote:
> > 
> > > There is however another problem that you haven't addressed yet, which
> > > is realtime. As far as I can tell, the realtime semantics require a
> > > strict ordering with respect to each other and their priorities.
> > > General approach can be either to limit all RT processes to a single CPU
> > > or, as we have done, declare a global RT runqueue.
> > 
> > Real time processes, when wakeup up fall calling reschedule_idle() that
> > will either find the CPU idle or will be reschedule due a favorable
> > preemption_goodness().
> > One of balancing scheme I'm using tries to distribute RT tasks evenly on
> > CPUs.
> > 
> 
> I think that would be a problem. My understanding is that if two RT process
> are globally runnable, then one must run the one with higher priority.
> Am I missing something here ?

It is not just the relative priorities of the realtime tasks, but
also the scheduling policy.  SCHED_FIFO (and to some extent SCHED_RR)
implies an ordering within the runqueue for tasks of the same priority.
This is difficult to achieve with multiple runqueues.  Most scheduler
implementations I am aware of, do something like what you suggested
above.

-- 
Mike