Ideally, there's (1) research for models; (2) making all relevant modifications to the p2p frameworks we choose to simulate (webtorrent should be the more straightforward for now, combining it with a DSN in the simulations would be a plus); (3) defining parameters; and (4) performing simulations along time, for different scenarios.
Goal: to have a reasonable estimate of network load (consumed bandwidth) according to growth rate and derived/other parameters.
An approach for torrent-based networks is described in items 3.0 e 3.1 here (along reference parameters): http://conferences.sigcomm.org/sigcomm/2004/papers/p444-qiu1.pdf
Ideally, there's (1) research for models; (2) making all relevant modifications to the p2p frameworks we choose to simulate (webtorrent should be the more straightforward for now, combining it with a DSN in the simulations would be a plus); (3) defining parameters; and (4) performing simulations along time, for different scenarios.
Section V in this paper presents a mould for such: http://www.netlab.tkk.fi/tutkimus/shok-fi/publ/p2pvodglobe.pdf
some previous incomplete tests can be seen here