Abstract: This paper discusses the primary concepts and methods related to distributed query processing. The paper also shows the characteristics of query using semi-join to fully reduce relations occurring in the query qualification and brings out a method of cost measurement Query processing algorithm based on semi-join is divided into three phases: copy identification phase, reduction phase and assembly phase. We also raise the characteristics of tree query and cyclic query, tree query recognition algorithm and transformation algorithm from cyclic query to tree query. Moreover, the paper approaches the optimal strategies of simple query and tree query, as well as heuristics algorithms based on semijoins.