We have created a new IPFS measurement report for calendar week 34 in 2023 from 2023-09-04 to 2023-09-11. You can find it here:
One slightly weird thing I’m seeing is that in the churn graph (https://github.com/plprobelab/network-measurements/blob/master/reports/2023/calendar-week-36/ipfs/README.md#churn) it looks like there are ~10k peers on kubo-v0.21 and ~3.5k on kubo-v0.18. But then in the agent version analysis (https://github.com/plprobelab/network-measurements/blob/master/reports/2023/calendar-week-36/ipfs/README.md#kubo) it looks like there are more than 20k peers on kubo-v0.18, which is also inline with what we see here: IPFS DHT | ProbeLab.
Am I missing something, or there’s something off here @dennis-tra ?
@yiannis that’s the number of “sessions”. A sessions represents the uptime of a peer with a start and end date. Peers that are online for the entire week (they successfully respond to every probe) don’t have an end date for their current session so they don’t appear in that Churn graph. This explains why we have 3.6k kubo 0.18 sessions in the churn graph but 20k 0.18 peers in the agents graph. Most of them are just very very stable.