[Nfd-dev] Help needed with ASF measurement

[Junxiao Shi]

Hi Ashlesh

This is probably a forwarding pipelines bug, but I won't have time to
diagnose this until at least Thursday.

Things to look at:

1. Find every code line that updates last renewed field.
2. Find which forwarding pipelines calls the function that updates last
renewed field.
3. Determine if there's a code path that generates an out record without
updating last renewed field.

Yours, Junxiao

On Oct 24, 2016 08:52, "Ashlesh Gawande (agawande)" <agawande at memphis.edu>
wrote:

Hi Junxiao


I am testing ASF in Mini-NDN.


The following line calculates the steady RTT by subtracting outRecord->
getLastRenewed()

from time::steady_clock::now()

https://github.com/named-data/NFD/blob/master/daemon/fw/asf-
measurements.cpp#L117

(which is called from beforeSatisfyInterest https://
github.com/named-data/NFD/blob/master/daemon/fw/asf-strategy.cpp#L115)


Sometimes the value of this steadyRTT is huge (ex: 200000 seconds) or
negative.

This makes ASF change faces to a higher RTT face.


So what could be the reason that this happening? I see much better and
expected results if I ignore these high/negative values.


(I see that https://github.com/named-data/NFD/blob/master/daemon/
table/pit-face-record.cpp#L34

initializes getLastRenewed to time::steady_clock::TimePoint::min(). So
maybe for some reason getLastRenewed does not get updated?)


Thanks

Ashlesh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/nfd-dev/attachments/20161024/8c83a400/attachment.html>