NS3 MPI : Questions regarding "nms-p2p-nix-distributed"

[dev]

Greetings,

I recently started using the distributed implementation of NS3 and have been trying to experiment with the examples included in the NS3 distribution. I had a following questions regarding the example "nms-p2p-nix-distributed" :

1. From my understanding of the implementation, the college networks (CNs) are assigned to the mpi processes in a round robin fashion, i.e. if the 8 CNs (numbered 0 to 7) are simulated on 5 processes (rank 0 to 4) than, rank 0 would simulate CNs : 0 and 5, rank 1 would simulate : 1 and 6,  rank 2 would simulate : 2 and 7,  rank 3 would simulate : 3 and rank 4 would simulate : 4. Is this correct ?
2. Keeping everything else constant, and varying the number of mpi processes in a simulation run, does not affect the simulation or in other words, it does not change the work done by the simulation as a whole. Is this correct ?

Thanks,

Dev

[pdbarnes]

Hello Dev,

1. I don't have the code in front of me to check.

2. Yes, as long as the smallest channel latency is long compared to the packet rate. Any additional work is for synchronizing across MPI, which should be negligible at the small number of ranks you are using and reasonable channel latencies.

Peter

[dev]

Hi Peter,

Thanks for the reply and I apologies for the my delay in replying to your comment.

As I mentioned in my post, I am using an example from the NS3 distribution. The example file is located at "src/mpi/examples/nms-p2p-nix-distributed.cc" in the latest NS3 distribution.

Best,

Dev