[ndnSIM] Inconsistent number of sent packets
Smallcat
huyao0107 at gmail.com
Mon Apr 22 16:57:29 PDT 2013
Hi Alex,
Thanks for explanation. One more question about the timeout of an interest before retransmission? How much is the time? Could I modify it easily in ndnsim?
Regards,
huyao
在 2013-4-23，3:21，Alex Afanasyev <alexander.afanasyev at ucla.edu> 写道：
> Hi again,
>
> If in your scenario you don't need retransmission at all, then you can write a simple requester app, similar to this one http://ndnsim.net/applications.html#dumb-requester. There is no way (at least as of right now) to disable retransmissions.
>
> Packets column in rate tracer refers to estimated rate in terms of packets. If you need actual number of packets within last period, then there is PacketsRaw column for that.
>
> ---
> Alex
>
> On Apr 20, 2013, at 5:09 AM, yao hu <huyao0107 at gmail.com> wrote:
>
>> Hi Alex,
>>
>> Yes. In my scenario, the network configuration is the same as the example. I just applied my forwarding strategy into it making part of interests sent from face 0 and 1 simultaneously. So there is traffic on both interface 0 and 1.
>>
>> Thanks for your explanation about L3RateTracer. Just now I also found that the more I print logging information (including self-written logging information), the less "Tx Packets" in visualizer or "PacketRaw" of "OutInterests" in L3RateTracer is. As you said before, the missing Interests are from the Interest retransmission which is also deemed as new Interests in "Cbr". So is there a way to decrease the Interest retransmission (at best to 0) from the time viewpoint?
>>
>> Another question is about the value of "Packets" column in rate-trace.txt. Why its value is not an integer? It refers to the layer 1 (or layer 2) packets which is not a ndn packet?
>>
>> Thanks for your point-out for "DidSendOutInterest". I added m_outInterests() to my CustomStrategy::DidSendOutInterest. the OutInterests value works well.
>>
>> Regards,
>> huyao
>>
>>
>>
>> 2013/4/20 Alex Afanasyev <alexander.afanasyev at ucla.edu>
>> Hi huyao,
>>
>> I run the grid scenario in visualizer with displaying interface statistics and i got exactly 2000 for interface 0 and 0 for interface 1:
>>
>> <stats.png>
>>
>> Did you make any other modification to the scenario? I really wonder how can you get any traffic on interface 1 in this scenario, as there should not be anything at all...
>>
>> About L3RateTracer. The line "20 Node0 2 dev=local(2) InInterests" tells you rate/number of interests that were received by NDN stack from the application face. From the application point of view, these are outgoing interests, but from the NDN stack point of view (since L3 tracer presents NDN stack point of view), these are incoming. I still don't really understand why do you have traffic on both interfaces... I suppose this is something that your custom forwarding strategy is doing?
>>
>> Btw. If in your strategy you're not using "TrySendOutInterest" method of the base class, then to have proper L3 logging you need to call DidSendOutInterest after interest were successfully sent out. Right now I see inconsistencies in numbers reported by L3RateTracer: line "20 Node0 0 dev[0]=net(0,0-1) OutInterests 0 0 0 0" should be non-zero.
>>
>> ---
>> Alex
>>
>> On Apr 19, 2013, at 8:16 AM, yao hu <huyao0107 at gmail.com> wrote:
>>
>>> Hi Alex,
>>>
>>> "Tx Packets" is from the "interface statistics" (right click the node from the ns3 visualizer) of the node0 which is the data requester in this grid scenario. Just now I tried it again by setting interest frequency as 100 (for 20 seconds), still the "Tx Packets" is 1805 (interface 0) and 22 (interface 1). The missing "Tx Packets" (around 200) are from the interest retransmission? I also used the L3RateTracer following your instruction (as follows). The "PacketRaw" column of
>>> 20 Node0 2 dev=local(2) InInterests
>>> refers the number of outgoing interest packets? Then what is the difference between this "2000" and that "1805"?
>>>
>>> Thanks very much for your time!
>>>
>>> Time Node FaceId FaceDescr Type Packets Kilobytes PacketRaw KilobytesRaw
>>> 20 Node0 0 dev[0]=net(0,0-1) InInterests 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) OutInterests 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) DropInterests 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) InNacks 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) OutNacks 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) DropNacks 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) InData 71.72 74.0581 1793 1851.45
>>> 20 Node0 0 dev[0]=net(0,0-1) OutData 0 0 0 0
>>> 20 Node0 0 dev[0]=net(0,0-1) DropData 0.12 0.12375 3 3.09375
>>> 20 Node0 1 dev[0]=net(1,0-3) InInterests 0.04 0.00105469 1 0.0263672
>>> 20 Node0 1 dev[0]=net(1,0-3) OutInterests 0 0 0 0
>>> 20 Node0 1 dev[0]=net(1,0-3) DropInterests 0.04 0.00105469 1 0.0263672
>>> 20 Node0 1 dev[0]=net(1,0-3) InNacks 0 0 0 0
>>> 20 Node0 1 dev[0]=net(1,0-3) OutNacks 0 0 0 0
>>> 20 Node0 1 dev[0]=net(1,0-3) DropNacks 0 0 0 0
>>> 20 Node0 1 dev[0]=net(1,0-3) InData 0.84 0.865859 21 21.6465
>>> 20 Node0 1 dev[0]=net(1,0-3) OutData 0.04 0.0412891 1 1.03223
>>> 20 Node0 1 dev[0]=net(1,0-3) DropData 0.72 0.742109 18 18.5527
>>> 20 Node0 2 dev=local(2) InInterests 80 2.12891 2000 53.2227
>>> 20 Node0 2 dev=local(2) OutInterests 0 0 0 0
>>> 20 Node0 2 dev=local(2) DropInterests 0.2 0.00507813 5 0.126953
>>> 20 Node0 2 dev=local(2) InNacks 0 0 0 0
>>> 20 Node0 2 dev=local(2) OutNacks 0 0 0 0
>>> 20 Node0 2 dev=local(2) DropNacks 0 0 0 0
>>> 20 Node0 2 dev=local(2) InData 0 0 0 0
>>> 20 Node0 2 dev=local(2) OutData 71.72 74.0581 1793 1851.45
>>> 20 Node0 2 dev=local(2) DropData 0 0 0 0
>>>
>>>
>>> 2013/4/19 Alex Afanasyev <alexander.afanasyev at ucla.edu>
>>> Hi huyao,
>>>
>>> Hmm. I just run ndn-grid scenario with frequency 10, 100, and 1000 and in all of the cases I see that the consumer send out exactly the theoretical amount of interests within 20 seconds (200, 2000, and 20000), though the same consumer has received different number of Data packets back due to buffer overflows (199, 1989, 2342).
>>>
>>> How exactly did you measure number of Interests and I'm not very sure what is "Tx Packets" you were referring to? I obtained my data using ndn::L3RateTracer. I put the following before Simulation::Run() in ndn-grid scenario (assuming the total simulation time is 20 seconds):
>>>
>>> boost::tuple< boost::shared_ptr<std::ostream>, std::list<Ptr<ndn::L3RateTracer> > >
>>> rateTracers = ndn::L3RateTracer::InstallAll ("rate-trace.txt", Seconds (19.99999999));
>>>
>>> (don't forget to include the header file:
>>> #include <ns3/ndnSIM/utils/tracers/ndn-l3-rate-tracer.h>
>>> )
>>>
>>> ---
>>> Alex
>>>
>>> On Apr 18, 2013, at 7:07 AM, yao hu <huyao0107 at gmail.com> wrote:
>>>
>>>> Hi Alex,
>>>>
>>>> The scenario I tested in is just the 9-node grid topology. I think maybe it has something wrong with my modified algorithm, since it shows no problem with a certain interest frequency, while with larger interest frequency (generally speaking) , the missing interest packets become obvious. I will check it soon. Thanks for your reply about routingCost. I see. :)
>>>>
>>>> Regards,
>>>> huyao
>>>>
>>>>
>>>>
>>>> 2013/4/18 Alex Afanasyev <alexander.afanasyev at ucla.edu>
>>>> Can you send out scenario (or link to scenario), so I can try to debug the problem on my machine?
>>>>
>>>> routingCost is basically an aggregated cost for the specific path. If you're using globalRoutingController, it sums up all the specified metrics on the path. For example, if you have this topology
>>>>
>>>> node1 ------ node2 ------- node3 ------- node4
>>>>
>>>> and all metrics for each link is 10, then routing cost on node1 to forward towards node4 using the only available face would be 30.
>>>>
>>>> If you are using "EnableDefaultRoute" in ndn::StackHelper, then this value doesn't really mean anything and suggests only that this "default" is the least preferred option to forward interests.
>>>>
>>>> ---
>>>> Alex
>>>>
>>>> On Apr 17, 2013, at 8:36 AM, yao hu <huyao0107 at gmail.com> wrote:
>>>>
>>>>> Hi Alex,
>>>>>
>>>>> I did not specify any randomization using Randomize parameter. I am not sure whether all the interests are consumed at the first attempt, but if that is a reason, the retransmitted Interest packets are not counted into "Tx Packets"? Besides, I am not sure about the meaning of routingCost column (2147483647) in FIB. Does it has something different from the 'metric' which affects the forwarding strategy to select the BestRoute? Could I specify it somewhere?
>>>>>
>>>>> Sorry for many questions. Thanks very much!
>>>>>
>>>>> Regards,
>>>>> huyao
>>>>>
>>>>>
>>>>>
>>>>> 2013/4/17 Alex Afanasyev <alexander.afanasyev at ucla.edu>
>>>>> Hi huyao,
>>>>>
>>>>> Did you specify any randomization for inter-interest gap using Randomize parameter (uniform, exponential)? Though, by default there is no randomization.
>>>>>
>>>>> Btw. Do all your interest are getting satisfied from the first attempt? If not, the difference can come from the interest retransmission, since those that needs to be retransmitted are sharing the same "cbr" constraints as new interests.
>>>>>
>>>>> ---
>>>>> Alex
>>>>>
>>>>> On Apr 16, 2013, at 8:32 AM, yao hu <huyao0107 at gmail.com> wrote:
>>>>>
>>>>> > Hi Alex,
>>>>> >
>>>>> > Just now, in order to verify my modification I used CustomerCbr to produce a certain number of Interest from a specific consumer, however, I found that there is a little difference between expectation and simulation result. For example, I set the Interest frequency as 10 (packets per second) for 20 seconds, the result is the number of sent packets is 200 which is the same as theoretic value. But if I set the Interest frequency as 100 for 20 seconds, the simulation result is the number of sent packets is 1805 which is less than the theoretic value 2000. What is the possible reason for this phenomenon? (No packets are sent from other faces)
>>>>> >
>>>>> > Thanks for your reply!
>>>>> >
>>>>> > Regards,
>>>>> > huyao
>>>>> > _______________________________________________
>>>>> > ndnSIM mailing list
>>>>> > ndnSIM at lists.cs.ucla.edu
>>>>> > http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> ndnSIM mailing list
>>>>> ndnSIM at lists.cs.ucla.edu
>>>>> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim
>>>>
>>>>
>>>
>>>
>>> _______________________________________________
>>> ndnSIM mailing list
>>> ndnSIM at lists.cs.ucla.edu
>>> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim
>>
>>
>> _______________________________________________
>> ndnSIM mailing list
>> ndnSIM at lists.cs.ucla.edu
>> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndnsim/attachments/20130423/074f6194/attachment.html>
More information about the ndnSIM
mailing list