Database replication

Abstract

A fault-tolerant node for synchronous heterogeneous database replication and a method for performing a synchronous heterogenous database replication at such a node are provided. A processor executes a computer program to generate a series of database transactions to be carried out at the fault-tolerant node. The fault-tolerant node comprises at least two relational database management systems, each of which are different relational database management system products, each implementing snapshot isolation between concurrent transactions. Each system comprises a database and a database management component. For each database transaction, operation instructions are provided concurrently to each of the systems to carry out operations on their respective databases and to provide respective responses. The responses generated by the systems either comprise an operation result or an exception. Only one of the systems is configured with a NOWAIT exception function enabled, which returns an exception when it is detected that two or more concurrent transactions are attempting to modify the same data item. The other systems are configured with the NOWAIT exception function disabled. The fault-tolerant node detects that two or more concurrent transactions are attempting to modify the same data item, and ensures that all systems apply the same order of modification of the data item by the concurrent transactions.