AW: [certi-dev] Question about NextEventRequest Service

[Roth, Christoph]

Hi,

while testing the cvs version I had problems with initializing several 
federates. The problem can be reproduced with the billard example. I started 4 
federates like this

rtig
billard -n1 -fTest -Ftest.fed
billard -n2 -fTest -Ftest.fed
billard -n3 -fTest -Ftest.fed
billard -n4 -fTest -Ftest.fed

in 4 different shells. When the simulatin is executed I get the following error 
during startup:

CERTI Billard 3.4.0cvs
with TIMESTAMP. If you want without TIMESTAMP add -e option.
HLALOG - 1288855783.143640 LibRTI::UnjoinedFederate - 
/home/christoph/software/certi_cvs/certi/libRTI/RTItypes.cc> RTI::Exception 
[FederationExecutionAlreadyExists,0x0] - reason=Test
BILLARD Note : FederationExecutionAlreadyExists Reason is : Test. OK I can join 
it
Display(400, 145, 500, 100)
Synchronization...
the class of the new created object is4
Declaration done.
RTIA:: RTIA has thrown ObjectNotKnown exception.
RTIA:: Reason: Object <1>not found in map set.

RTIA: Statistics (processed messages)
List of federate initiated services 
--------------------------------------------------
       1 Message::OPEN_CONNEXION (MSG#1)
       1 Message::JOIN_FEDERATION_EXECUTION (MSG#5)
       1 Message::SYNCHRONIZATION_POINT_ACHIEVED (MSG#11)
       1 Message::PUBLISH_OBJECT_CLASS (MSG#29)
       1 Message::PUBLISH_INTERACTION_CLASS (MSG#31)
       1 Message::SUBSCRIBE_OBJECT_CLASS_ATTRIBUTES (MSG#33)
       1 Message::SUBSCRIBE_INTERACTION_CLASS (MSG#35)
       1 Message::REGISTER_OBJECT_INSTANCE (MSG#41)
       1 Message::ENABLE_TIME_REGULATION (MSG#81)
       1 Message::ENABLE_TIME_CONSTRAINED (MSG#83)
       2 Message::QUERY_FEDERATE_TIME (MSG#86)
       1 Message::TIME_ADVANCE_REQUEST (MSG#92)
       2 Message::GET_OBJECT_CLASS_HANDLE (MSG#113)
       3 Message::GET_ATTRIBUTE_HANDLE (MSG#115)
       1 Message::GET_INTERACTION_CLASS_HANDLE (MSG#117)
       3 Message::GET_PARAMETER_HANDLE (MSG#119)
       1 Message::GET_OBJECT_CLASS (MSG#128)
  150043 Message::TICK_REQUEST (MSG#142)
       4 Message::TICK_REQUEST_NEXT (MSG#143)

List of RTI initiated services 
--------------------------------------------------
       3 NetworkMessage::MESSAGE_NULL (MSG#2)
       3 NetworkMessage::SET_TIME_REGULATING (MSG#7)
       1 NetworkMessage::TIME_REGULATION_ENABLED (MSG#9)
       1 NetworkMessage::TIME_CONSTRAINED_ENABLED (MSG#10)
       1 NetworkMessage::ANNOUNCE_SYNCHRONIZATION_POINT (MSG#13)
       1 NetworkMessage::FEDERATION_SYNCHRONIZED (MSG#15)
       3 NetworkMessage::DISCOVER_OBJECT (MSG#43)

 Number of Federate messages : 150070
 Number of RTIG messages : 13
UN Socket(RecevoirUN) : : Connection reset by peer
HLALOG - 1288855805.990042 LibRTI::2 - 
/home/christoph/software/certi_cvs/certi/libRTI/RTItypes.cc> RTI::Exception 
[RTIinternalError,0x0] - reason=NetworkError in tick() while receiving 
response: Error while receiving UN message.
address@hidden:/home/christoph#  TCP Socket  3 : total =       490 Bytes sent 
 TCP Socket  3 : total =      1513 Bytes received
 UDP Socket  5 : total =         0 Bytes sent 
 UDP Socket  5 : total =         0 Bytes received


The federate in which the error occurs differs. Sometimes it occurs in several 
federates. Do you have any idea?

Thanks and regards,
Christoph




-----Ursprüngliche Nachricht-----
Von: address@hidden [mailto:address@hidden Im Auftrag von Eric Noulard
Gesendet: Mittwoch, 3. November 2010 08:48
An: CERTI development discussions
Betreff: Re: [certi-dev] Question about NextEventRequest Service

2010/11/3 Roth, Christoph <address@hidden>:
>> I mean, it **should** be ok, but since the design is very new and
>> there is no other equivalent
>
> What about other commercial RTI implementations (e.g. pRTI). Are they not 
> able to cope with the time creep problem until now?

I don't know because I'm not using those :-]
More seriously, AFAIK
time creep problem has been addressed in two major different ways using iether
 optimistic time advancing technique like time warp, or some kind of
Mattern "distributed snapshots",...
The algorithm proposed in CERTI is not optimistic (like time warp) and
is somehow different from Mattern snapshot.
We (Pierre and me) present the CERTI algorithm in more detail some day.

Concerning others,
I don't know which algorithms are implemented in Picth  or MAK or any
other commercial RTI.

That said the best way to go would probably be to test drive several
RTI (Open Source or commercial)
and see how they compare wrt this kind of issue, may be your test case
will help for that.

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

-- 
CERTI-Devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/certi-devel