Streaming Raft algorithm implementation in kotlin

Raft is a consensus algorithm created by Diego Ongaro in an attempt to replace Paxos with a simpler and more understandable consensus algorithm.

The original raft paper can be found here.

The idea behind this project, is to create a streaming raft implementation written in kotlin, by using the concept of a sliding unacked data window. This means that the leader will NOT wait for every follower ACK before sending more data, it will just keep pushing data until the window is full or the ACK is received releasing space from the in-flight data window.

This is a personal pet project, so it is still lacking automated tests, comments and overall documentation. This might be fixed in the future, if I have time.

Back in 2015, while working at PaddyPower Betfair there was a need to create a streaming platform that could replace kafka and be resilient to network partitions, as it was proven that kafka could not handle network partitions very well.

After some due dilligence, the team decided to write their own in-house implementation of the Raft algorithm, with a few slight modifications to allow increased performance without sacrificing the consistency.