> For networks larger than ~250 nodes, single-host simulation becomes > unreliable regardless of available RAM.
I can probably ask for access to a 64-core server, in case that helps. Please let me know. > --------------- > Our initial tests with babeld show: > Grid 100 nodes: 100% connectivity, ~14s convergence > Chain 50 nodes: 100% connectivity, ~5s convergence > Small-world 100 nodes: 100% connectivity, ~12s convergence By default, Babel uses a hello interval of 4s. Your results are between 1 and 3.5 hello inervals, and I suspect that you're measuring the performance of the link quality estimator: the link quality estimator needs 3 hello intervals to converge, and the protocol converges almost immediately afterwards. In order to test this hypothesis, please try reducing Babel's hello interval (to 1s, then to 0.5s, then to 0.1s, then to 0.05 s). As long as the convergence time is on the order of 3 hello intervals, you're measuring the performance of the link quality estimator. Once the convergence time is significantly higher than 3 hello intervals, you're measuring the performance of the routing protocol itself. -- Juliusz _______________________________________________ Babel-users mailing list [email protected] https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users
