Scalability

[brice-morin]

It seems that this new group works very well at the beginning, then the adaptation times tend to grow, from 200ms at the beginning to 5-10 seconds when we start to have more than 100 nodes. Also I suspect that funny things are happening like nodes considered as left (because they do not answer to ping or because master do not have time to handle their pong) whereas they are still there.

Attached is a log of the whole simulation (I started nodes 25 by 25, waiting in between (longer and longer) that the master is "available" again).

logs.txt

[maxleiko]

Could you give me the big fat model generated at the end with your hundreds of nodes?

[brice-morin]

Not that big... 3.7 MB. Just rename into a JSON file.

model509.txt