How the simulation is done

nWaku Version: v0.34.0-rc1

Looks like the message delay is not consistent among previous simulations.

The intention of this report is to check if the length of a simulation has an impact on message propagation delay.

The setup is composed of 3 bootstrap nodes, 500 relay nodes and a consistent rate of 1 message per second and 1KB. The only difference is the number of the messages (thus, length of the simulation).

The injection is done as following:

1. A message injector container selects randomly one node from the mesh.
2. Request this node to publish a message through Rest API
3. Repeat steps 1 and 2 over the duration of the simulation each message rate time (once per second).

Results:

Initially, we tried 75, 150, 200, 300, 600 messages.

• 75 messages
• 150 messages
• 200 messages
• 300 messages
• 600 messages

With 300 and 600 messages, there are some messages that takes up to almost 2 seconds.

We decided to repeat the 300 messages simulation, and the results went back to normal: