Construction of Multiple-Catalog Database for JVO

We present a development of an efficient query system for the Multiple Catalog Database. In a case where a VO user collects all the available information about an astronomical object, it is necessary to send queries to all the VO services. However, this would be inefficient since most services tend not to have data on the target object. We thus developed a new query system for users to efficiently find all the available information from multiple catalogs. This system collects basic data (coordinates and fluxes) of astronomical objects from the VO data services and major astronomical catalogs, and store them into a local database built with PostgreSQL. The design of this database is one of key issues; an efficient query mechanism for more than billions of objects is required. We employed the Table-Partitioning technique and developed a method to build queries for the partitioned tables. We compared our method with the partitioning function of PostgreSQL and found that our method is more efficient by a factor of 7 to 150. We discuss the architecture of this system.