Complex Methods and Class Allocation in Distributed OODBSs

Abstract

In a distributed object-oriented database system (DODBS), queries which invoke methods executing at different sites and access different classes need to be executed very efficiently. Therefore, the methods invoked and classes accessed by the queries need to be allocated to sites so as to reduce the data transfer cost in processing a given set of queries. The methods and class allocation(MCA) problem needs to take into consideration complex interdependencies among queries, methods and classes. In this paper, we develop a comprehensive cost model for total data transfer incurred in processing a given set of queries by incorporating the dependencies among the queries, methods and classes. Further, we develop an iterative approach to generate near-optimal solution for the combined MCA problems by using the above cost model. In this approach, we start with an initial class allocation(CA) which is used for method allocation(MA), which in turn is used for CA, and so on. We stop this iterative MCA when there is no further reduction in total data transfer cost incurred in processing the given set of queries. We also present the results of experiments conducted to evaluate the effectiveness of our approach by comparing the results with exhaustive enumeration solution (which guarantees the optimal solution).