[ndnSIM] Issue with RttEstimator.

Saeid Montazeri saeid.montazeri at gmail.com
Sun May 18 09:51:38 PDT 2014


Hi all,

I found an strange behaviour with ndnSIM related to RttEstimator. I made a
very simple scenario to debug. There are five nodes in a linear topology.
Suppose one consumer node is connected to router_0 and one producer node is
connected to router_4.  There are two cases:
1) the consumer node installs CcnxConsumerCbr to request the content on the
producer with rate of *1 *CcnxConsumerCbr per second and the simulation
time is *10* second.

2) tthe consumer node installs CcnxConsumerCbr to request the content on
the producer with rate of *0.1 *CcnxConsumerCbr per second and the
simulation time is *100* second.

The execution times are very different. That is, case 2 has longer
execution time. After digging dipper, I found that the number of call for
RttEstimator:RetransmitTimeout makes the difference.

RetransmitTimeout is called by CcnxConsumer::CheckRetxTimeout which
periodically calls itself with the period of m_retxTimer with the default
value of 50ms.

This makes a real issue for the scenarios with large number of
CcnxConsumerin terms of execution time. That is, this leads to a lot
of unnecessary
events. In addition, I found that for my evaluation which is network of
caches performance (network with rare congestion), it is not necessary to
have m_retxTimer very small. I changed it to 500ms and the execution time
drastically is decreased.

Please let me know your comments, specially about the right value of m_
retxTimer for different scenarios. I think the smaller the
m_retxTimerleads to faster detection of the network
RTT fluctuation by consumers.

In addition, do you think that I need to change the default value of m_
minRto and m_maxRto?

Best Regards,
Saeid
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndnsim/attachments/20140519/bacd2205/attachment.html>


More information about the ndnSIM mailing list