Hi. Am 06.03.2009 um 16:12 schrieb Szabolcs Nováczki:
Not yet but sounds like a good idea, but how can you measure this.
I once wrote a little test application that did nothing more than sending floods of UDP packets to a destination. We recorded the incoming UDP packets at the destination with tcpdump and analyzed the resulting race with wireshark later. The disruption shows in a number of missing packets (gap) in the trace. We also tried iperf for generating the flood of packets but it had problems when switching IP addresses and stopped sending when the handoff occured. I don't know if there is a smarter way to do it (there probably is) but we got quite convincing results with this method.
I can send you the flooder program if you whish.
Could you send me the trace once again (on private - maybe as zip or tar). Somehow I did not get the attachment.Could you have a look at the trace?
The 2.13GHz Pentium with 2G RAM laptop we use as mobile should be fast enough, I guess.
Sure. Crypto should only be a minor factor then. BR, Tobias
I hope this provides some insights into the problem. BR, Tobias Am 06.03.2009 um 14:03 schrieb Szabolcs Nováczki: Hi!Yes, I did not get your message on the list. Wonder why... Anyway. Here is the answer: See my answers inline! ################# Hi,I will try do give some assistance but I need a bit more information aboutthe problem.a) Could you provide some numbers in ms or s? What means significant? Iwasnot able to read your attachment and it probably lacks timing informationanyway.Sorry i forgot to give this info :). The delay is about 2 secs. I attachedthe file again but you have right: there is no timing.b) Could you also give some detail about the machines you are working with(CPU speed) ... Ill send this later cause i am not sure of the detailes.c) ... and the test setup you are using (directly connected vs. sited atdifferent sites).All is IPv6. There are three different /64 networks. One for the cn andtwofor the mobile. The cn is attached on wire. The mobile is attched to oneof the latter two sites through wireless acces points and uses IPv6 statelessautoconfiguration when receiving router advertisements to configure its IP address. We use a script which periodically changes (iwconfig new_ap) from one ap to the other and back. The mobile reconfigures its ip address whichtriggers the hip update mechanism. d) Do you move between different (wireless?) networks? see answer for c.) e) Are you using the same IP address in these different networks? We use different addresses: 2001:0738:2001:2084:0213:ceff:fe7b:fd28 2001:0738:2001:2088:0213:ceff:fe7b:fd28 BR, Tobias ############## Br, Szabi 2009/3/6 Tobias Heer <tobias.heer@xxxxxx> Just in case you missed the message on the list...Anfang der weitergeleiteten E-Mail: Von: Tobias Heer <heer@xxxxxxxxxxxxxxxxx>Datum: 6. März 2009 09:59:05 MEZ An: "hipl-users@xxxxxxxxxxxxx" <hipl-users@xxxxxxxxxxxxx> Kopie: Goodzi <goodzi@xxxxxx>Betreff: Re: [hipl-users] long delay in mobility handling procedureHi,I will try do give some assistance but I need a bit more informationabout the problem.a) Could you provide some numbers in ms or s? What means significant? Iwas not able to read your attachment and it probably lacks timing information anyway.b) Could you also give some detail about the machines you are workingwith (CPU speed) ...c) ... and the test setup you are using (directly connected vs. sited atdifferent sites). d) Do you move between different (wireless?) networks? e) Are you using the same IP address in these different networks? BR, Tobias Am 06.03.2009 um 09:11 schrieb Szabolcs Nováczki: Hi hipl users!We are doing some tests with infraHIP recently. With basic mobilitytests we run into the following behaviour:There is a significant delay in infraHIP (on the mobile side) beforesending out the first update package.I attached a commented trace of the hipd output (hipd_trace_medium)where I highlighted the suspected part.My request is that if some of you has time to analyze this or someone knows the reason for this behaviour please share with us! What happenesthere? Is this delay ok? We are also interested of mechanisms triggering the update procedure in hipd, and how this trigger is processed. Thx! Br, Szabolcs NovaczkiPs.: hipd is running on 2.6.27-11 generic kernel installed from ubuntuinterpid package-- Dipl.-Inform. Tobias Heer, Ph.D. Student Distributed Systems Group RWTH Aachen University, Germany tel: +49 241 80 207 76 web: http://ds.cs.rwth-aachen.de/members/heer<hipd_trace_medium.txt>-- Dipl.-Inform. Tobias Heer, Ph.D. Student Distributed Systems Group RWTH Aachen University, Germany tel: +49 241 80 207 76 web: http://ds.cs.rwth-aachen.de/members/heer
--
Dipl.-Inform. Tobias Heer, Ph.D. Student Distributed Systems Group RWTH Aachen University, Germany tel: +49 241 80 207 76 web: http://ds.cs.rwth-aachen.de/members/heer