Maintaining consistency of replicated data in multidatabase systems

The paper presents two protocols for maintaining consistency of replicated data in multidatabase systems. The protocols meet both autonomy and consistency requirements by employing different replica control and commitment approaches: unilateral local commitment and deferred propagation for local applications, and two-phase commitment and immediate propagation for global applications. The first protocol ensures replication consistency (with regard to one copy serializability) through the use of a global certification protocol. The second protocol, which is an extension of the first by using propagation locks on primary copies, allows consistency certification to be performed locally for global queries that only read data copies from a single site. We identify and examine the major issues relevant to the proposed protocols.