Re: [lmap] Feedback on draft-eardley-lmap-terminology

[Paul Aitken <[e-mail]>]

Juergen,

> On Wed, Jul 24, 2013 at 09:18:07PM +0100, Paul Aitken wrote:
>   
>> MA's should do what they're requested by controllers as best as
>> they're able. I can't immediately think of a situation where an MA
>> could be given conflicting instructions. However, if conflicts are
>> possible the MA could enqueue the requested tests to be run
>> sequentially to avoid inter-test conflict. If conflict can arise
>> from two controllers, then it can also arise by requesting the same
>> tests from a single controller - so designing this limitation into
>> the system gains nothing.
> C1: do test X
> C2: don't do test X

You're supposing inter-controller communication, such that C2 knows that 
C1 requested X ?

The MA should do test X as requested by C1. It should ignore the command 
from C2, because C2 didn't request test X.

It's clearer if the sequence was:

C2: do test X
C1: do test X
C2: don't do test X


Then C2 is cancelling its previous command, whereas C1's command stands.


> or
>
> C1: do test X every 10 minutes
> C2: do test X every 8 minutes
>
> What should the MA do?

The MA should do test X multiple times exactly as requested.

If both requests are received at exactly the same instant (which is a 
corner case, even if we're talking about multiple processors + multiple 
interfaces), then every 40 minutes test X could be run just once as an 
optimisation.

More realistically, test X will be run at t0, t10, t20, t30, ... for C1 
and at t1, t9, t17, t25, ... for C2. These times never coincide.

P.