Re: [certi-dev] NPMA

[Eric Noulard]

2012/1/19 Roth, Christoph (ITIV) <address@hidden>:
> Hi,
>
> I just had a look into the paper “Design and modeling techniques for
> real-time RTI time management” and was wondering if the NPMA can be denoted
> as either synchronous or a asynchronous algorithm.
>
> As far as I understood, the NMA is asynchronous since no global barriers are
> generated and a federate is allowed to advance as soon as it is able to
> calculate a local LBTS.
>
> In case of the NPMA you say, you do not have to look for an appropriate
> instant to start a wave/reduction. So, I assume the NPMA is also an
> asynchronous algorithm that additionally exploits global information?

Yes that's the case.
You can think of NPMA as way to do a reduction of LBTS using the collected
NMA sent by some federates at any instant.

Due to current CERTI architecture where all NMA messages flows through
a central point (RTIG),
as soons as the RTIG finds a better LBTS it sends "spontaneous"
classical NULL message
to the concerned federates which [safely] boost them.

To summarize:
1) NPMA is both conservative AND asynchronous.
2) It is easily implemented in CERTI thanks to current CERTI architecture.
3) It should theoretically be possible to implement it in a
distributed (still asynchronous)
manner.

If you successfully use NPMA for your testcase and/or have some more
technical question
I'd be glad to discuss it with you.

-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org