[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[certi-dev] Re: CERTI-Devel Digest, Vol 55, Issue 11
From: |
Christian Stenzel |
Subject: |
[certi-dev] Re: CERTI-Devel Digest, Vol 55, Issue 11 |
Date: |
Fri, 29 Oct 2010 21:53:33 +0200 |
Hello,
Pierre and Eric are experts here, but I suppose to know, that CERTI uses a
null-message synchronization scheme.
What you observed is known throughout the literature as lookahead creep.
You can learn more about this phenomena on these slides:
http://www.cc.gatech.edu/classes/AY2006/cs4230_fall/Lectures/06-Lookahead%20(9-14-05).ppt
In short: To synchronize different federates or so-called logical processes you
can use null message synchronization (also known as chandy, misra, bryant
algorithm). A nullmessage is an explicit guarantee that a federate which is
able to schedule events for another federate, will never create events with a
timestamp smaller than its own time plus it's own lookahead.
In a classical distributed discrete event simulation, a federate (or logical
process) would send a nullmessage after each time advance. The receiver of that
nullmessage can advance its local time to the timestamp of the received
nullmessage. If the lookahead is relatively small the whole distributed
simulation advances it's time only in small portions. When uisng nullmessage
for synchronization you can not advance to the time of the next event directly.
So, I think that everything is okay. No failure at your side.
Christian
-------- Original-Nachricht --------
> Datum: Fri, 29 Oct 2010 12:01:08 -0400
> Von: address@hidden
> An: address@hidden
> Betreff: CERTI-Devel Digest, Vol 55, Issue 11
> Send CERTI-Devel mailing list submissions to
> address@hidden
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://lists.nongnu.org/mailman/listinfo/certi-devel
> or, via email, send a message with subject or body 'help' to
> address@hidden
>
> You can reach the person managing the list at
> address@hidden
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of CERTI-Devel digest..."
>
>
> Today's Topics:
>
> 1. Question about NextEventRequest Service (Roth, Christoph)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Fri, 29 Oct 2010 16:21:23 +0200
> From: "Roth, Christoph" <address@hidden>
> Subject: [certi-dev] Question about NextEventRequest Service
> To: <address@hidden>
> Message-ID:
> <address@hidden>
>
> Content-Type: text/plain; charset="us-ascii"
>
> Hi,
>
>
>
> I have got a question concerning synchronization. My scenario is the
> following:
>
>
>
> I have two federates f1 and f2. Both are event based and use the
> NextEventRequest (NERQ) Service to advance time. The simulation loop in
> both federates looks like this:
>
>
>
> ...
>
> rtiAmb->updateAttributeValues();
>
>
>
> rtiAmb->nextEventRequest(newTime);
>
> while(isAdvancing) {
>
> rtiAmb->tick2();
>
> }
>
> ...
>
>
>
> Both federates have a lookahead of 10. Assume both federates are at the
> same time (=0). Now, if both federates request for the same new time
> with very high value like newTime=1000000 the tick2()-call lasts very
> long (several seconds) until the simulation proceeds. Have you any idea
> about the reason for this long delay? It seems as if the delay depends
> on the low lookahead, because if I increase the lookahead up to the
> newTime value simulation is much faster.
>
>
>
> In my opinion the low lookahead should have no influence on execution
> speed since both federates guarantee to not send any message with a
> lower timestamp than newTime+lookahead, so the simulation should be able
> to directly advance (This is what I could read in the IEEE 1516.1.2000
> standard page139), or did I make a misstake?
>
>
>
> My setup:
>
> - Ubuntu linux
>
> - Local execution
>
> - Certi 3.4.0
>
>
>
>
>
> Thank you very much and regards,
>
> Christoph
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> http://lists.nongnu.org/archive/html/certi-devel/attachments/20101029/a9648bcb/attachment.html
>
> ------------------------------
>
> _______________________________________________
> CERTI-Devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/certi-devel
>
>
> End of CERTI-Devel Digest, Vol 55, Issue 11
> *******************************************
--
GMX DSL Doppel-Flat ab 19,99 €/mtl.! Jetzt auch mit
gratis Notebook-Flat! http://portal.gmx.net/de/go/dsl
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [certi-dev] Re: CERTI-Devel Digest, Vol 55, Issue 11,
Christian Stenzel <=