Spread Concepts - Replicated Hash Table

Behind the scenes of the Replicated Hash Table (RHT), each local RHT table is identical to the corresponding RHT table replicas on other participating systems. This allows developers to distribute queries to any of the table replicas, thereby linearly increasing the query throughput of the entire system with each additional replicated table.

Updates to any of the replicas are applied to all the replicas in the exact same manner on all of the machines. This ensures that perfect consistency of the replicas is maintained at all times.

RHT uses innovative algorithms from Spread Concepts to maximize scalability and performance without sacrificing consistency guarantees. Our solution scales to tens of replicas either in a cluster or over the Internet. Each of the table replicas is considered a master table, in comparison to the traditional single master / multiple slave architecture. Updates between the replicas are propagated and ordered with real-time performance.

The overall replication architecture is very efficient. High performance is achieved because:

End-to-end acknowledgments are not needed on a regular basis. They are only used when the connectivity of the network changes.
Highly efficient network dissemination and message ordering is gained by using The Spread Toolkit.