Rafka: P2P mesh for distributed messaging

Infrastructure

Rafka is an experimental Rust message broker. It replaces the standard leader-follower coordination model with a P2P mesh architecture. It uses a DHT for metadata management and an in-memory sidecar database to reduce storage latency. Moving away from a centralized coordinator is an interesting architectural choice. However, P2P routing for data streams often trades simplicity for complex consistency issues. I would like to see the sample sizes and benchmarks used to justify this shift over traditional models. Specifically, how does the DHT perform under high node churn? It is a useful resource for those exploring decentralized stream routing, provided you account for the experimental nature of the implementation.

Source

Rafka: A P2P-mesh approach to distributed messaging