sguzwf/chdb

A simple distributed database using Cloud Haskell

chdb is a simple distributed document storage system. It was written as a toy to play with Cloud Haskell. The system provides methods to store documents, and handles conflicts through simple multi version concurrency control. The goal is to provide high availability, concurrency, and fault tolerance.

Master

There is a single master server. The master spawns all slaves and directs client messages to a slave capable of handling the client's request. The master is also responsible for directing replication of documents.

Slaves

The slave processes handle all the data, they are distributed on nodes throughout the network. Clients communicate directly with the slaves, after an initial exchange with the server.

Startup

The master starts, and spawns all the slave nodes and processes. When the slaves start they send a message to the master with a list of the DocumentIds (and the versions of those Documents) that the slave has. The server builds and maintains a mapping of DocumentIds (and versions) to slave Pids (OR SendPorts!). The master accepts the first response for a given DocumentId, until a newer version of that document is discovered. If a newer version of a Document is discovered, The new version is made the working copy and the old version is the backup. The Document is queued for replication.