Processing queries in distributed systems based on local area networks include local processing for the queries and transmission of relations or attributes to remote sites. To reduce the query processing time, semijoin operations can be used for reducing the size of relations to be transmitted. In the literature, semijoin of a relation $R_1$ with other relations, $R_2$,$R_3$,$\ldots$,$R_n$, on a particular joint attribute is achieved by performing the following consecutive semijoins: semijoin of $R_2$ to $R_1$, semijoin of $R_3$ to $R_1$,$\ldots$, semijoin of $R_n$ to $R_1$.
In this thesis, a method for semijoins is proposed by using intersections of the joining attributes, and the performance of the proposed method is studied. With respect to transmission time, the transmission time of the proposed method is equal to that of the traditional method above, when the order of transmission is the same in two method. However, with respect to the local processing time, when the size of a relation is twice greater than the size of the joining attribute, the proposed method has less local processing time than that of the sequential method. Only when the selectivities of the joining attributes are small, for example less than 0.5, the traditional method outperforms the proposed method with respect to the local processing time. This thesis also proposed query optimization algorithms using the proposed semijoin method for the two types of queries: simple queries and general queries.
