Optimization techniques are essential for the efficient query processing in object databases, mainly for the evaluation of path expressions. Such expressions represent a path defined over a sequence of relationships between collections of objects and it has been the focus of numerous researches associated to eminent technologies, such as object-relational databases and the XML language. Most works found in the literature about query optimization considers that cost metrics are used to guide the search of execution plans with good performance.

    This thesis proposes a cost model that allows the estimation of the performance of various strategies for the evaluation of path expressions over distributed-object databases. Metrics were specified to represent the cost of data input/output operations (I/O), CPU instructions, and communication between two nodes. The wideness of the proposed model distinguishes this thesis from other works, since it covers a high number of relevant aspects of (i) the OO data model, such as the selectivity of a path expression and partial participation of the collections in the relationships, (ii) the model of data storage according to diverse policies of object clustering, (iii) the execution model of algorithms, and (iv) the collections distribution project, considering different techniques of fragmentation and the fragments allocation.

    The modeling of aspects related to the impact of object distribution on the evaluation of path expressions is the major contribution presented in this dissertation, which aims to accomplish a great hiatus left due to the absence of related works.