Wishlist of graphs or questions we want to answer with the data

[dennis-tra]

We have a list of graphs from our preliminary analysis here:

https://www.notion.so/pl-strflt/NAT-Hole-punching-Success-Rate-2022-09-08-Data-Analysis-a67ab92efd6e43259da28a3c000a559d

---

After discussing them we came up with these other perspectives we want view our data:

• Hole Punch Attempts split between transports
• Matrix of Continent Relaying Peer X Continent Punchr Client
• Success rate stability for remote peers (if we hole punch the same peer multiple times, do we ALWAYS succeed with the hole punch or only sometimes?)
• Is TCP more timing sensitive
  • maybe TCP as successful but we drop because QUIC is faster

[mxinden]

Low priority:

• By default go-libp2p aims to listen via 2 relay nodes. I wonder whether these values are good as is, too high or too low. Questions could be: (a) how many relayed addresses do nodes advertise and (b) how many of the advertised relays are reachable?

[dennis-tra]

Covered Success rate stability for remote peers (if we hole punch the same peer multiple times, do we ALWAYS succeed with the hole punch or only sometimes?) here.

[dennis-tra]

Not sure how to find out Hole Punch Attempts split between transports. When we receive the CONNECT message we see all the advertised addresses from the remote peer. We try to dial all of them simultaneously I believe. Then we end up with a direct connection on one of the advertised transports.

Not sure how to split the analysis by transport in this case :/

[mxinden]

In the case of TCP and QUIC, we could disable one locally and thus test the other.