MVME177 CPU Utilization

MVME177 CPU Utilization

[Ron Thompson]

Howdy few, brave.  I asked before and got a few opinions via the mail,
and I really appreciate hearing from the only bunch I care to hear from.
I'll ask again, just for yuks, to see what the few and the brave Ada
folks think...

In order to measure CPU Utilization, it is proposed that the figures
for two 177s, two that are blissfully unaware of one another, be
averaged together in order to meet required overhead reserves.
Any opinions???  The email is fine for responses if you care to, and
Thank You all.
-- 

rct

The opinions above are mine and mine alone.

Re: MVME177 CPU Utilization

[Steve O'Neill]

Ron Thompson wrote:
> In order to measure CPU Utilization, it is proposed that the figures
> for two 177s, two that are blissfully unaware of one another, be
> averaged together in order to meet required overhead reserves.
> Any opinions???  The email is fine for responses if you care to, and
> Thank You all.
> --
> 
> rct
> 
> The opinions above are mine and mine alone.

That's not a valid measure of system reserves.  Unless, of course, one
processor can borrow time from the other or allocate any overload work
to the other processor.  My approach would be to take the worst case
loading of the two processors.

Of course this is only my opinion.

Steve O'Neill

Re: MVME177 CPU Utilization

[Markus Nickels]

Steve O'Neill wrote:

> Ron Thompson wrote:
> > In order to measure CPU Utilization, it is proposed that the figures
>
> > for two 177s, two that are blissfully unaware of one another, be
> > averaged together in order to meet required overhead reserves.
> > Any opinions???  The email is fine for responses if you care to, and
>
> > Thank You all.
> > --
> >
> > rct
> >
> > The opinions above are mine and mine alone.
>
> That's not a valid measure of system reserves.  Unless, of course, one
>
> processor can borrow time from the other or allocate any overload work
>
> to the other processor.  My approach would be to take the worst case
> loading of the two processors.
>
> Of course this is only my opinion.
>
> Steve O'Neill

I agree to Steve. Your approach would be acceptable (from a customers
point of view) if you have a mechanism which allocates dynamically (at
run-time) processing power between the processors. If you choose a
software architecture which gives you the possibility to distribute the
processes statically (at link-time) this may be acceptable, too, but
depends on your customer requirements.
But you pointed out that the processors don't know anything about each
other, so your approach won't be accepted.
It may be interesting how you measure the workload of your processors.
Additionally I'm interested in the characteristics of your project
(cyclic/acyclic processing, hard/soft real time conditions, worst case
scenarios etc.) because I'm working with similar problems.

Markus Nickels
Markus.Nickels@t-online.de

Re: MVME177 CPU Utilization

[Ron Thompson]

Markus Nickels wrote:
> I agree to Steve. Your approach would be acceptable (from a customers
> point of view) if you have a mechanism which allocates dynamically (at
> run-time) processing power between the processors. If you choose a
....
> Markus Nickels
> Markus.Nickels@t-online.de

If email ok Markus, drop me a note and we can continue.  Thanks for your
response in front of the
group.                                                           
                                                                         
                                                                       
-- 
                                                                       
rct
                                                                        
The opinions above are mine and mine alone.