🌌 IPFS Measurement Report - 2023 Week 10 - 2023-03-12

We have put together a new IPFS measurement report for calendar week 10 in 2023 from 2023-03-06 to 2023-03-13 . You can find it here:

:sparkles: Notable things:

  • We have new DHT performance graphs - median provider record lookup performance in eu-central-1 is ~200ms
  • Rotating Nodes continue to decrease - which is a good thing :+1:
  • We don’t report website performances this time but will do next week again. We restructured our measurement setup and are much more confident about the numbers we’re gathering. However, we don’t have enough to report yet.
1 Like

The number of PeerIDs seen is dropping even more, but this is the result of less nodes rotating their PeerIDs. The network seems to be more stable than ever, with 50% of nodes being always online: network-measurements/README.md at master · protocol/network-measurements · GitHub.

This doesn’t translate to reduced churn though, which is a bit weird (?)

Great to see 200ms lookup performance from EU, although not clear why us-west-1 is not as good despite most nodes appearing to be in the US :thinking: Would it make sense to do a one-off experiment (say for 1 week) to some other location in the US to see what is the performance comparison?

This doesn’t translate to reduced churn though, which is a bit weird (?)

If these peers are not reachable at all they don’t count towards the churn. Let’s say these peers just appear in the DHT because they went online only for a very brief moment, then the crawler will find them in the DHT but won’t track an online session for them.

Great to see 200ms lookup performance from EU, although not clear why us-west-1 is not as good despite most nodes appearing to be in the US :thinking:

Maybe most of the peers are deployed in us-east-* and the transatlantic link is faster than the intercontinental one?

Yup, could be. I think it’s worth considering doing an experiment from another US location to figure out if this is the case.