An Analytical Approach Towards Evaluating Method-Induced Vertical Partitioning Algorithms

Vertical partitioning is a technique for reducing the number disk accesses to execute a query by minimizing the number of irrelevant instance variables accessed. The complexity of object-oriented database (OODBs) models due to subclass hierarchy and class composition hierarchy complicates the definition and representation of vertical partitioning of the classes. Furthermore, as the data in the OODB is directly accessed by the encapsulated methods, there is a need to obtain the cost of executing methods for evaluating the utility of vertical class partitioning. In this paper, we develop a cost model for method execution, introduce three algorithms to perform vertical class partitioning (namely, affinity-based, cost-based, and mixed), and show that the mixed approach is not only efficient but also provides mostly optimal (otherwise, near-optimal) vertical class partitioning scheme.