[cgl_discussion] requirements for configurable Round Robin quantum and 1 ms tick

[Eric.Chacron at alcatel.fr]

Hi Peter,

PRF 1.0 says "CGL shall provide the capability of configuring the scheduler
to provide soft real time support so that real time scheduling latency of a
given task will not exceed a target defined by the vendor. Based on
commodity hardware commonly supported by Linux, latency responses in the
range of 10 to 15 ms should be considered reasonnable and likely"

I agree my proposal fit in PRF 1.0, but it is very more specific as it
applies to Round Robin policy only and ask for
a configurable maximum quantum allowing to switch a task and queue it at
the end of the list of task competing with the same priority
when it has kept the processor for more than the quantum. The second part
is to allow 1 ms tick on Intel architecture.

So , i agree it can be presented as a subrequirement of PFR 1.0 provided we
update it
(is it still possible to update 2.0 the req. spec. ? ) or then integrated
as a new one in 3.0.

PS/ i think that maybe detailed requirements have more chance to be
accepted and implemented as they
can be simply understood without any confusion.


Eric





Peter Badovinatz <tabmowzo at us.ibm.com>@westrelay02.boulder.ibm.com on
08/12/2003 07:13:14 PM

Sent by:    wombat at westrelay02.boulder.ibm.com


To:    Eric CHACRON/FR/ALCATEL at ALCATEL
cc:    cgl_discussion at osdl.org
Subject:    Re: [cgl_discussion] requirements for configurable Round Robin
       quantum  and 1 ms tick


Eric.Chacron at alcatel.fr wrote:
>

Eric,

I've been traveling quite a bit during and since Ottawa, am traveling
again.  So, I actually missed this (it was buried in my in-box).
Apologies on that.

> Today some applications need to use Real Time process with Round Robin
> policy.
>
> Today the quantum seems to be defined by a constant value as 20 x the
tick
> value.
> Then if i use a 10ms tick the quantum is equal to 200 ms. This seems too
> large.
>
> A second requirement could be to enable a 1 ms tick period for the timer
> interrupt on Intel Pentium architectures.
> Is'it compatible with kernel maximum latency ? ( remenber that some
> application runs whithout any disk and without
> related sources of latency like fork+ exec ...)
>
> I would like to have the feed-back from everyone who's interrested in the
> question and also from Montavista
> as ther could be some link with low latency and RT scheduler for the
second
> point.

According to the "rules" from Ottawa, we aren't intending to add any
additional requirements.  Very specifically, no priority ones.  That
said, we do have an open "soft real time" (PRF.1.0) v2.0 requirement
which rolls together the latency and premption requirements.

What I'm wondering about is if your requests could be included with
PRF.1.0?  The question is if these are already possible and potentially
supported, I have no problem adding them.  If these require additional
work and don't easily fit into the current work, we may be out of luck
for v2.0.

I'll echo your comment, we need folks with more knowledge of scheduling
and real time to step in.

Peter
--
Peter R. Badovinatz aka 'Wombat' -- IBM Linux Technology Center
preferred: tabmowzo at us.ibm.com / alternate: wombat at us.ibm.com
 These are my opinions and absolutely not official opinions of IBM, Corp.