[ndnSIM] Number of dropped packets

[Alex Afanasyev]

Hi Mohammad,

I found several inconsistencies in drop tracing (in forwarding strategy and l2-rate-tracer).
In particular, l2-rate-tracer was recording only "rates" which are calculated using exponential smoothing averaging, this could be one reason for large discrepancies (in case you compared absolute numbers of drops versus rates).

Another inconsistency is related to what was traced in ndn::ForwardingStrategy.  As of the latest commit, I have changed the logic to the following:

- every time forwarding strategy picks a face to forward (no matter of the actual status of sending out of a particular face), this even is traced as OutInterest or OutData.

- if status of the send was negative, this will be reported in a separate trace via DropInterest and DropData.

These DropInterest and DropData are now outputted in ndn::L3RateTracer and ndn::L3AggregateTracer.

Can you check the updated code?

---
Alex


On Mar 18, 2013, at 8:56 AM, Hovaidi Ardestani Mohammad <mohammad.hovaidi.ardestani at aalto.fi> wrote:

> Hi Alex
> Thank you for response!
> The problem is that each consumer is expressing unique interest every time and there should not be any caching in the path, but interestingly, the discrepancy was large enough to convince me to ask this question.
>  
>  
> From: Alexander Afanasyev [mailto:cawka1 at gmail.com] On Behalf Of Alex Afanasyev
> Sent: Monday, March 18, 2013 4:44 PM
> To: Hovaidi Ardestani Mohammad
> Cc: ndnsim at lists.cs.ucla.edu
> Subject: Re: [ndnSIM] Number of dropped packets
>  
> L2Tracer (there is no ndn namespace) shows number of packets dropped due to queue overflows.  In many cases of normal NDN communication, number of incoming interests is not equal to number of outgoing interests (same for data), as NDN router (in general) performing much more operations than just forwarding packets.
>  
> For example, if the same data piece is requested from two different neighbors, only the first interest will be forwarded, the second one will be either "collapsed" or satisfied from caches.   When Data packet comes, if there is no corresponding PIT entry (or entries), Data packet will be dropped (without L2Tracer knowing it).  If there is a PIT entry with multiple incoming faces (or multiple PIT entries), then more than one Data packet will be forwarded out.
>  
> At the same time, if you have a simple simulation where consumers requests unique pieces of content, then your formula should be correct, unless you're using forwarding strategy that enforces interest limits (e.g., PerOutFaceLimits).  How big is the number discrepancy?
>  
> ---
> Alex
>  
> On Mar 18, 2013, at 2:41 AM, Hovaidi Ardestani Mohammad <mohammad.hovaidi.ardestani at aalto.fi> wrote:
> 
> 
> Hello everybody!
> I am thinking that the following formula should give us the number of dropped packets.
> Number of Dropped packets = (Number of received interests + number of received data packets) - (Number of sent interests + number of sent data packets).
> If above formula is correct, then why on earth number of dropped packets shown by ndn::L2RateTracer is not equal with that?
> Thank you for your response in advance!
> -Mohammad
>  
> _______________________________________________
> 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/20130318/1a359bb9/attachment.html>