Introducing Khepri for Metadata Storage in RabbitMQ

Michael Davis from the RabbitMQ team presents Khepri, a tree-like, replicated database library that utilizes the Raft distributed consensus algorithm. Aiming to replace Mnesia for persistent and replicated metadata storage within RabbitMQ, Khepri boasts several features that make it particularly suitable for this role. Among its recent updates are a new 'projections' feature that enhances read speeds, an overhauled API, and planned integration into the RabbitMQ codebase under a feature flag. Davis discusses the implementation details, usage examples, performance improvements, and the decision-making process that led to the development of Khepri.