RandyKV is a distributed, in-memory key-value database that utilizes gossip-based communication and randomized hashing for effective load balancing.
RandyKV is built on top of a gossip-based communication protocol. As such, it requires nodes that exchange data with aid from a node discovery service. In RandyKV, there is a single type of node, called an agent, that is able to participate in the system by both sending (pushing) and receiving (pulling) data.
In order to participate in the system, an agent must communicate with the node discovery service, which is a backend service that maintains global state of all participating nodes in the system.
- Agents, which maintain central database logic, are implemented using C++
- Node discovery, which maintains global state of all existing nodes, is implemented in TypeScript (NestJS), using Google's gRPC.