Constant experiment parameters
Notes:
IDONTWANT was introduced on nim-libp2p 1.2. We can see the effect of IDONTWANT control messages in the 50KB and 500KB plots, where the bandwidth from 1.2 to onwards gets decreased with respect to 1.1.
Bandwidth incrase in 1.6 is due to a Kubernetes worker misconfiguration.
Quic looks like cannot handle the connectivity of the nodes, and the mesh is not stable. The bandwidth is lower because there are a steadily increasing number of nodes that have less than d_low peers, making the publishing not working. We saw a behavior like this before, but we thought it was related to the artificial delays we were introducing in the experiments. In this ones, we removed artificial delays for quic, but looks like there is still a problem.
Quic metrics
Size: 1KB, Mesh: 1000 nodes
In the case of 1KB, v1.11.0 performs exactly the same as other versions (excluding quic).