Material: Health Dashboard | Performance Dashboard | GitHub Repository | libp2p Discourse

https://youtu.be/fyhZWlDbcyM

Context

The libp2p team is rolling out a NAT hole-punching solution, which makes use of the Direct Connection Upgrade through Relay (DCUtR) protocol. The target of this project is to measure the success rate of libp2p’s NAT hole-punching.

<aside> 🔥 Impact: Figuring out the success rate and potential improvements to the NAT Hole Punching technique proposed by libp2p will help us overcome a mighty problem in P2P networks in general, and IPFS in particular.

</aside>

Progress

The base methodology is given below. For more details on the methodology, the infrastructure set up, as well as the project progress, please check the Github repository linked above. Meeting notes from our sync meetings (and not only) are given further down.

<aside> 🛠 We track open ToDo’s in GitHub issues.

</aside>

Analysis

NAT Hole-punching Success Rate - 2022-09-08 Data Analysis

NAT Hole-punching Success Rate - 2022-09-29 Data Analysis

NAT Hole Punching Measurement Analysis 2022-12-08

NAT Hole Punching Measurement Analysis 2022-12-15 Mother of All Demo Days

Measurement Campaign Analysis Notes (WIP)

Final Report: NAT Hole Punching Measurement Campaign (draft)

Methodology

• Development
  1. Honeypot libp2p host
     • participates in the DHT
     • allow tons of incoming connections
     • enumerates the DHT to come across many/all peers and hope that they put us in their routing table
     • Persist events of incoming connections into DB (multiaddresses + peer ID)
  2. Thin API layer
     • allows querying a p2p-circuit multi addresses that are capable of hole punching
     • allows persisting hole punch results
     • gRPC for easy client generation in Go+Rust
     • encapsulates logic of whom to serve which Multi address to hole punch
  3. Puncher Client (Go + Rust)
     • queries API for multi address to hole punch
     • initiates hole punch
     • reports back the result
• Deployment
  • postgres as data persistence layer
  • Deploy Honeypot + API Layer on VM (separate processes)
  • Install Grafana + Prometheus to monitor these processes
• Metrics
  • Success Rate
  • Duration
  • Attempt Count

Punchr Client registration