Scalable Relational Database Replication

A method of replicating a relational database comprising:

• receiving a plurality of database data modifications from a plurality of clients in parallel at a primary replication agent of a primary database;
• providing the plurality of database data modifications in parallel from the primary replication agent of a primary database to a secondary replication agent of a secondary database;
• applying the plurality of database data modifications from a plurality of clients in parallel to the secondary database;
• serially receiving a plurality of prepare statements in sequence for the plurality of database data modifications from the plurality of clients at the primary replication agent of the primary database;
• serially forwarding the plurality of prepare statements in sequence for the plurality of database data modifications from the primary replication agent to the secondary replication agent;
• and
• applying a plurality of database data modifications and commit statements in parallel in prepare sequence to the secondary database to maintain a sequence of database data modifications in the secondary database.

A system for scalable replication of a relational database comprising:

• a first independent shard database comprising a first shard primary database, a first shard primary replication agent, a first shard secondary database, and a first shard secondary replication agent;
• and
• a second independent shard database comprising a second shard primary database, a second shard primary replication agent, a second shard secondary database, and a second shard secondary replication agent,
• wherein the first shard secondary replication agent controls replication of the first shard primary database to the first shard secondary database independently of the second shard secondary replication agent controlling replication of the second shard primary database to the second shard secondary database.

A method for failover of a first database of a replicated relational database comprising:

• starting a failover replication agent of a failover database;
• informing a primary replication agent of the failover replication agent;
• receiving transactions at the failover replication agent;
• rebuilding a new instance of the first database at the failover database under the control of the failover replication agent while the failover replication agent is receiving the transactions without further system interruption.