Maintaining Consistency of Interdependent Data in Multidatabase Systems

Interdependent data are characterized by dependency constraints and mutual consistency requirements. Maintaining consistency of interdependent data that are managed by heterogeneous and autonomous DBMSs is a real problem faced in many practical computing environments. Supporting a mutual consistency criterion that is weaker than one copy serializability is often acceptable if better performance can be achieved and the autonomy of DBMSs is not sacrificed. Updates to interdependent data have to be controlled in order to maintain their consistency. We propose a solution where at least one of the copies of each interdependent data, called current copy, is kept up-to-date in the system. Using the concept of update through current copy, we show how a weaker mutual consistency requirement, called eventual consistency, can be satisfied. The proposed approach requires writing only the local copy (as opposed to writing many or all copies) and hence preserves autonomy by not requiring synchronization of the local concurrency controllers. This approach exploits the semantics of both the interdependent data and transactions; thus, it is non-intrusive, flexible and efficient.