The majority of currently commercially available database management systems enable just two basic replication mechanisms: asynchronous, as an one-way or two-way replication, and synchronous, usually realised through ROWA protocol. Generally, application of the synchronous protocol causes a hardly repairable data inconsistencies. While using the ROWA protocol, unavailability of any node within the system completely impedes the execution of update transactions, which essentially reduces the data availability. There are no satisfactory replication mechanisms at hand if data availability and data consistency are similarly important. Based on a two-tier replication protocol, a new replication protocol has been designed in order to enable adaptation of the consistency level and data availability to a current availability of nodes within the system. An extended data dictionary has been designed and implemented, containing the fragmentation and allocation metadata and the criteria to control the consistency and availability of data. The generator of the triggers and stored procedures source code, using the extended data dictionary data, enables orchestrated action of the new replication mechanism together with the ROWA protocol. It offers the possibility to control the data consistency and availability in a wide range, from a high availability and low consistency levels to a complete and simultaneous data consistency in all the system nodes, which is paid by reduced data availability. The described replication model is suitable for implementation using mechanisms present in the majority of modern commercial systems, it requires only minimum changes in the application layer and it has demonstrated good properties while decentralising an existent complex information system.