[Nfd-dev] NameTree entries

[Dehart, John]

On May 7, 2017, at 12:33 PM, Junxiao Shi <shijunxiao at email.arizona.edu<mailto:shijunxiao at email.arizona.edu>> wrote:

Hi John

Yes, this makes sense. Sending Interests with 10 random components at the end would make NameTree to have 10x entries as PIT entry.

I’m not sure I quite understand ths. Lets say I have 5 Interests:

/abc/zyx/1234
/abc/zyx/2345
/abc/zyx/3456
/abc/zyx/4567
/abc/zyx/5678

And lets say they are all still in the PIT, unsatisfied and have not timed out.

How many NameTree entries would you expect there to be?


On the oher hand, PIT entry count seems high. If the node isn't currently handling heavy traffic, this indicates a previous Interest flooding attack in which InterestLifetime was set to a large value.

That is exactly the kind of extreme case we did. It was a bit of an accident but it is what ended up happening.
We were trying to load the CS with pre-recorded packets. So, we used tcpreplay to send a batch of interests with
long lifetimes, then used tcpreplay to send a batch of matching data.

Thanks,
John




Yours, Junxiao

On Sun, May 7, 2017 at 09:28 Dehart, John <jdd at wustl.edu<mailto:jdd at wustl.edu>> wrote:

We also have an extreme example:

     nNameTreeEntries=1673990
           nFibEntries=3
           nPitEntries=167400
  nMeasurementsEntries=0
            nCsEntries=252277

where there are on the order of 10 times more name tree entries than PIT entries.

Does that make sense?

John


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/nfd-dev/attachments/20170507/a6472550/attachment.html>