Lehrgebiet Informationssysteme

A multidatabase consists of a collection of autonomous local databases. Systems used to manage multidatabases are called multidatabase systems (MDBSs). In such a system, global transactions are executed under the control of the MDBS. Independently, local transactions are submitted directly to a local DBS (LDBS) by local applications. An MDBS should provide a mechanism to globally manage transactions. However, global transactions are long-living and involve operations on multiple and autonomous local databases. Moreover, MDBSs do not have any information about the existence and execution order of local transactions. Thus, conventional approaches to manage transactions are unsuitable for MDBSs. In this paper, we address the reliability problem in MDBSs. For this purpose, we propose two types of protocols for making MDBSs resilient to failures. One type of protocol should enforce that, when a given global transaction completes its execution, it has the same state (committed or aborted) at every site it has run. The other type determines the actions to be triggered after failures in a multidatabase environment. These protocols can reduce the frequency of global transaction undo after the occurrence of failures, and make the MDBS able to deal with failures which may occur in a multidatabase environment.

in Proc. XIV Brazilian Symposium on Databases (SBBD 1999), Florianopolis, Santa Catarina, Oct. 1999.