Consistency Control of Replicated Data in Federated Databases

to whether a promise can be issued is based on a satissability test, that the constraints associated with those transactions already in the queue, together with the constraint associated with the new transaction, are satissable. The constraint solver is also capable of answering queries about the ealiest time at which a piece of data will become up-to-date. The transaction mechanism supports the lter-ing and alerting of transactions. Filtering applies to transactions which cannot satisfy the consistency requirements. As an extreme example, the system might not accept a transaction updating the primary copy if the computer storing the secondary copy is under a maintenance scheduled to last for more than one hour, and the connection is speciied with time limit set to one hour. Alerting applies to transactions which should operate at given levels of consistency. For example , assume that two copies are related by an identity connection with a time limit of two days; then, a transaction that needs reading the latest value should not operate on the secondary copy. Alerters could give signiicant messages to users, such as: \The data will not be available until a given time." 4 Conclusion The eeective use of distributed databases requires a proper balance between autonomy and coordination. A critical decision innuencing the autonomy of local sites concerns the method for update propagation to replicated data. We have designed an identity connection model to specify replication, permissible delay in the consistency of copies, and data derivation rules. This model allows us to declare the nature of the coupling between autonomous sites in a distributed database. We are developing a transaction processing mechanism for consistency control of replicated data based on the identity connection model. The mechanism uniies existing approaches to update propagation, and supports the ltering of update transactions that cannot meet consistency requirements, as well as the alerting of query transactions that cannot see consistent data. The users of modern information systems are faced with autonomous distributed data. Our work provides a general framework for the large amount of research that addresses problems related directly or indirectly to the issue of dealing with replicated data. Our approach for specifying consistency requirements of replicated or derived data in a network of autonomous, federated databases is based on a new concept: the identity connection1]. An identity connection links attributes of two, not necessarily distinct, relations or fragments of relations , generally allocated at …