Re: periods and deadlines in SCHED_DEADLINE

[Luca Abeni <lucabe72@email.it>]

Hi all,

first of all, thanks for including me in these emails, and sorry for the
delay...

On Fri, 2010-07-09 at 15:38 +0200, Raistlin wrote:
> Hi all,
> 
> So, talking to Peter and Thomas at the OSPERT workshop in Brussels [1],
> the so called "sporaidic task model" came out many many times!
I assume here you are simply talking about tasks with relative deadline
different from period, right? (the term "sporadic" is more often
associated with non-periodic activation patterns).

[...]
>  - do you think it could be useful to have a different syscall to deal
>    with the period parameter (if it's different from deadline), e.g.,
>    something useful to make the task periodic as you have (if I remember
>    well) in Xenomai or RTAI?
Maybe I am confused because I missed the initial part of the discussion,
but here I think there is the risk to mix two different concepts: the
"reservation period" (that is, the period P used to postpone the
scheduling deadline when the budget arrives to 0), and the "task
period" (which has to do with the periodicity of tasks activations). For
implementing a periodic behaviour in the task (this is, AFAIK, what RTAI
similar API provide), no new syscall is needed: clock_nanosleep() is
enough. See http://www.disi.unitn.it/~abeni/RTOS/rtapi.pdf for a
(stupid) example.
The reservation period, on the other hand, is a scheduling parameter,
and I think that setting it with extended versions of sched_setparam(),
sched_setscheduler() and similar is ok.


>    If you think it's worth doing that, do you think the
>    task_wait_interval() syscall that we already added could/should do
>    the job?
I do not remember what task_wait_interval() does :)
Is it the syscall you added to indicate the end of a job?


> Basically, from the scheduling point of view, what it could happen is
> that I'm still _NOT_ going to allow a task with runtime Q_i, deadline
> D_i and period P_i to use more bandwidth than Q_i/P_i, I'm still using D
> for scheduling but the passing of the simple in-kernel admission test
> Sum_i(Q_i/P_i)<1 won't guarantee that the task will always finish its
> jobs before D.
I think if you want a different P_i and D_i you can use D_i for
generating new scheduling deadlines on task arrivals as "d = t + D_i",
and P_i to postpone the scheduling deadlines as "d = d + T_i" when the
budget is 0.
Depending on the replenishment amount you use, you might need to modify
the admission test as "Sum_i(Q_i/min{P_i,D_i}) < 1" or not (if you
always replenish to Q_i, then you need a modified admission test;
otherwise, you can compute the replenishment amount so that the
admission test is unaffected).


			Thanks,
				Luca