XML is a standard for exchanging and formatting data over the Internet and XQuery is a standard query language for searching and integrating XML data. Therefore, it is a natural choice for interoperability to use XQuery over the Internet. Global XQuery queries search and integrate heterogeneous data, being distributed in the local systems. In order to process efficiently global XQuery queries, their processing strategy is important because an improper processing strategy could produce an enormous number of intermediate results or execute redundant expressions. In distributed relational databases, there are some techniques for processing global SQL queries. Unfortunately, however, the structure of the data handled by the XQuery language is quite different from the one by the SQL. The XQuery language deals with semi-structural data, i.e. treestructured data, while SQL deals with well-structured data, i.e., the table-shaped data. These structural differences make it difficult to apply the techniques for global SQL queries into for global XQuery queries. Especially this paper considers the join cost for devising a query processing strategy. Therefore, we define some problems for estimating the join cost in XQuery queries and propose ECNJ algorithm for solving these problems. Also this paper proposes the query processing strategy and evaluates the strategy by implementing a prototype system.