The Congruity replication technology offers low latency, high throughput data store replication, supporting applications that require high availability and high performance.

The Congruity Replication Engine enables transparent Multi-Master or Peer Replication. Unlike traditional master/slave replication architectures, each replica serves as a master. Actions can be initiated from any replica, and are immediately replicated to all the other participating replicas. The engine guarantees that all actions are applied at all the replicas in the exact same order, and thus that all replicas remain identical. The engine is highly scalable and supports tens of replicas either in a cluster or over the Internet.

Even in local area networks, network partitions occur regularly due to either hardware (e.g. temporarily disconnected switches) or software (heavily loaded servers). In wide area networks, partitions can be even more common. The Congruity Replication Engine can recover from a wide range of network and server failures. It operates correctly even in environments that experience server crashes and network partitions. When partitioned components remerge, the engines exchange information about actions that are known to each component, This exchange enables all components to share the same knowledge and the data to appear identical on both sites

The Congruity Replication Engine utilizes advanced and provably correct protocols and methods to efficiently generate and disseminate a global persistent consistent order of all actions. Congruity Engine can sustain many hundreds of transactions per second.

Several elements contribute to the Congruity Engine's high performance and scalability.

• Optimized synchronization algorithms minimize latency while providing high throughput.
• The Congruity algorithm avoids unnecessary delays resulting from end-to-end acknowledgments. Unlike typical two-phase commit mechanisms, the Congruity protocol does not require an end to end acknowledgement for every action. Instead, end-to-end acknowledgements are used only once for every network connectivity change event (such as network partition or merge).
• Read only queries can be performed locally, and need not be sent over the network, thus providing high performance and availability.
• Synchronous disk writes are almost eliminated, without compromising consistency.
• Highly efficient real time ordered local and wide area network dissemination is gained by using the Spread group communication toolkit.
• The Congruity architecture eliminates the problem of conflicts between replicas because updates are always invoked in the same order at all the replicas, and therefore no conflicts are created.

The replication engine supports various semantic models, from strict consistency and fully ordered actions, to weaker models, relaxing or enforcing the consistency constraints as needed by the application.

Congruity is a generic replication engine which runs outside the replicated data store and can be seamlessly integrated with existing databases and applications. The engine can even replicate a combination of data stores from different vendors.