Max: I assume this is a placeholder, but in case it's not, just wanted to make sure it's known that the red dots don't match where us-west-2 and us-east-1 are.

Reworked the map. Thanks!

Not sure if this appropriate to add here, but if this is informing about how we think about latency, three things I think we always need to do on reporting latency: 1. Be very clear on when the timer starts and when it ends 2. Show percentiles (not average) 3. Show count (so can make judgement on how representative the percentile metrics are). I assume there are great blog posts out there on "the why" for those things above, but they are tenets I would want us to hold to when we discuss latency.

I would think "ms" is a better unit than "s" here (and below) for anything that is subsecond.

I find seconds, more generally SI units, easier to reason about.

Per above, whenever we talk about percentile I think we also need to provide count to provide context.

What do the dotted lines mean in the diagram?

Not sure what ChatGPT had in mind there. Replaced by solid lines.

I believe this was addressed https://research.protocol.ai/blog/2020/honey-i-shrunk-our-libp2p-streams/ but regardless, is not directly related to this post in my opinion.