INTERACTION BETWEEN CONCURRENT TRANSACTIONS IN THE DESIGN OF DISTRIBUTED DATABASES

A distributed database system allows concurrent execution of transactions from multiple users at multiple locations. This paper presents a general scheme of including the effect of interaction between such transactions in the design process of distributed databases. This scheme allows the decision maker to choose the most appropriate design for a distributed database. The interaction effect is a consequence of concurrency control procedures and contention for system resources. Prior research has not included all components of the interaction effect in the design process. A separate module, called an interaction module, was built to measure the interaction effect using a comprehensive simulation model of a general purpose relational distributed database management system (DBMS). The interaction module can be interfaced with any distributed database design system as an add-on module so that the interaction effect can be included in the design process. A detailed example demonstrates the use of the interaction module in the decision-making process. Experiments conducted using the module show that the interaction effect can be significant and that it can affect the choice of the design. Experiments were also conducted to measure the sensitivity of design choices under changing input conditions.