Executing multidatabase transactions

In a multidatabase environment, the traditional transaction model has been found to be too restrictive. Therefore, several extended transaction models have been proposed in which some of the requirements of transaction, such as isolation or atomicity, are optional. The authors describe one of such extensions, the flexible transaction model and discuss the scheduling of transactions involving multiple autonomous database systems managed by heterogeneous DBMS. The scheduling algorithm for flexible transactions is implemented using L.0, a logically parallel language which provides a framework for concisely specifying the multidatabase transactions and for scheduling them. The key aspects of a flexible transaction specification, such as subtransaction execution dependencies and transaction success criteria, can be naturally represented in L.0. Furthermore, scheduling in L.0 achieves maximal parallelism allowed by the specifications of transactions, which results in the improvement of their response times. To provide access to multiple heterogeneous hardware and software systems, they use the Distributed Operation Language (DOL). DOL approach is based on providing a common communication and data exchange protocol and uses local access managers to protect the autonomy of member software systems. When L.0 determines that a subtransaction is ready to execute, it hands it through an interface to the DOL system for execution. The interface between L.0 and DOL provides the former with the execution status of subtransactions.<
>