[ndnSIM] Realistic but small topology for cache performance evaluation

[Saeid Montazeri]

Dear Alex,

Thanks a lot for the reply. The problem with rocketfuel is that the number
of routers is more than 1000 as you know. I simulated an unrealistic
topology which is attached with only 10 core routers and 8 edge routers.
One simulation run for 2000 sec took 6 hours. You can imagine what the
execution time of a topology with 1000 routers will be.

Using MPI is an interesting idea. However, I am not sure how much it can
help. It is due to the fact that in my experiences the execution time
exponentially increases with the number of node. If the execution time
decreases linearly with the number of nodes (nodes=cores for the best case)
by using MPI, I think it may not help a lot. This is only a guess and I am
going to test it.

Best Regards,
Saeid

On Wed, May 29, 2013 at 2:34 PM, Alex Afanasyev <
alexander.afanasyev at ucla.edu> wrote:

> Hi Saeid,
>
> I don't really have an answer to your question... You could try to use
> something like rocketfuel ISP maps (
> http://www.cs.washington.edu/research/networking/rocketfuel/maps/rocketfuel_maps_cch.tar.gz)
> for two different ISPs and then connect them in some way (here
> https://github.com/cawka/ndnSIM-sample-topologies are versions converted
> into more ndnSIM-friendly format).  It is not exactly "realistic" and I
> hope other have better suggestions here, but at least the scale is
> reasonable (e.g., AT&T topology is about 700 or 800 nodes).
>
> Another possibility to handle long simulation times for large topologies
> is to try to use MPI module.  For this, you would need to partition the
> topology in some meaningful way (if you can afford one node per CPU, that
> would be the extreme case).  I did some successful experiments with a
> simple topology and manual topology partitioning, with reduction of
> simulation time almost proportionally to the number of CPUs used.
>
> ---
> Alex
>
> On May 28, 2013, at 9:06 PM, Saeid Montazeri <saeid.montazeri at gmail.com>
> wrote:
>
> Dears in ndnSIM,
>
> I want to evaluate the cache performance in NDN. I need a realistic
> topology. I have seen some topology generators such as inet. In addition, I
> investigated some traces such as those measured by rocketfuel. However, the
> problem with these tools and traces is the topology scale. That is, the
> topology has a large number of nodes which make the execution time of the
> simulation very long. In addition, I need a topology which has at least two
> connected ISPs because I want to measure the exchanged traffic between
> ISPs.
>
> Please let me know if you have any suggestion.
>
> Best Regards,
> Saeid
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndnsim/attachments/20130530/60a1b5c3/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: meshtopo.eps
Type: application/postscript
Size: 560730 bytes
Desc: not available
URL: <http://www.lists.cs.ucla.edu/pipermail/ndnsim/attachments/20130530/60a1b5c3/attachment.eps>