An approach to query processing in object oriented distributed database systems is proposed. Distributed PIOS is a server that supports an object-oriented data model, physical data independence (i.e. different strategies for storing class hierarchies, grouping, horizontal and vertical partitioning of objects), and fragmentation transparency (i.e. transactions are not aware of the distribution of database fragments on several nodes of a computer network). The problem of the optimization of distributed queries (i.e. determining which data must be accessed at which site and which data must be transmitted among sites) is the focus of the paper.