The major difference of heterogeneous distributed database systems (HDDBSs) from traditional homogeneous distributed database systems is the local site autonomy which permits local database system (LDBS) to retain its own design and its own decision. Global serializability can never be sacrificed at all in HDDBSs. To guarantee the global serializability of the concurrent execution of global transactions and local ones, the existing LDBSs need to be integrated in a tightly coupled way so that the global serializability can be preserved even when global update operations are allowed.
The transaction management in tightly coupled HDDBSs is difficult problem since the local site autonomy as well as the global serializability must be preserved. However, most of previous concurrency control schemes fail to preserve the global serializability which is a strict correctness criteria for concurrency control. Also, these fail to preserve the local site autonomy.
We propose a unified concurrency control scheme that guarantees both global serializability and local site autonomy in tightly coupled HDDBSs in which two different types of concurrency control schemes are used as local concurrency control scheme:two-phase locking scheme and timestamp ordering scheme. By unified, we mean that the feature of two-phase locking scheme and that of timestamp ordering scheme are integrated into the proposed scheme. Besides the advantage of preserving both global serializability and local site autonomy, the unified concurrency control scheme is able to maintain the original degree of local concurrency of LDBS in which a timestamp ordering schemlme is used as local concurrency control scheme since the unified scheme schedules transactions according to the type of local concurrency control scheme.
To resolve the global deadlock problem which can occur in HDDBSs even when every LDBS already has the different local deadlock resolution scheme, among the deadlock resolution schemes which...