NS-3 can support very large number of nodes. However, your computer will become the bottleneck and you will have to go to a distributed system.
pdbarnes
unread,
Nov 13, 2013, 2:07:53 PM11/13/13
Reply to author
Sign in to reply to author
Forward
Sign in to forward
Delete
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to ns-3-...@googlegroups.com
In [1] we used plain vanilla ns-3 and fit 750K nodes in 32 GB. This tells you the gross memory scaling for ns-3 (~42kB/node, including code and data, for mixed CSMA and P2P channels, YMMV).
Since each MPI rank has to have the entire model, even in a distributed (parallel) model, you are limited by the amount of memory on a single compute node. The ARL work cited by Konstantinos went through some hoops to instantiate portions of the 360M node model on each distributed compute node (MPI rank), forming a sort of federated model.
We (LLNL) are working on changes to ns-3 that will break this memory barrier, allowing each MPI rank to instantiate only a portion of the model. This will enable much larger models, limited by your cluster memory, not your compute node memory.