A star schema is very popular for modeling data warehouses and data marts. Therefore, it is important that a database system which is used for implementing such a data warehouse or data mart is able to efficiently handle operations on such a schema. In this paper we will describe how one of these operations, the join operation --- probably the most important operation --- is implemented in the IBM Informix Extended Parallel Server (XPS).

A database server used for implementing a data warehouse must support other features than a database server used for OLTP. Therefore, in this paper we will look specifically at features necessary for efficiently processing queries on a database with a star schema model, a database scheme which is used very often in data warehousing. We will especially analyze the features provided for this by the IBM Extended Parallel Server (XPS). There are special star join methods like the Push-Down Hash Semi Join, new access methos like Generalized Key (GK) indices, and specific index usages like multi-index scans for supporting the efficient processing of star schema queries.

In this paper, we investigate which problems exist in very large real databases and describe which mechanisms are provided by Informix Extended Parallel Server (XPS) for dealing with these problems. Currently the largest customer XPS database contains 27 TB of data. A database server that has to handle such an amount of data has to provide mechanisms which allow achieving adequate performance and easing the usability. We will present mechanisms which address both of these issues and illustrate them with examples from real customer systems.