Non-parametric estimation techniques in support of query decomposition strategies for heterogeneous distributed database management systems

Abstract

Organizations such as NASA have vast repositories of data stored in non-relational databases. In order to access information in these environments, the queries used need to be optimized over heterogeneous (relational and non-relational) distributed databases. This paper presents non-parametric estimation techniques in support of the query decomposition process for heterogeneous distributed database management systems (HDDBMSs). In particular, the approach that has been developed facilitates the size estimation of intermediate results when the underlying databases are very large and the majority of the queries are expressed over attributes that have a small set of unique values. Size estimation techniques are presented for the select, project, normal join and semijoin operations. An analysis of the size estimation approach for the join operation is presented also. These concepts and techniques are illustrated by presenting a heterogeneous sample astrophysics database stored at different NASA locations.