Attribute partitioning algorithm in DOODB

In distributed object-oriented databases (DOODB), objects are distributed in different sites on communication networks. In DOODB, class fragmentation to divide a class into fragments is needed for improving performance and for reducing data transfer. Class fragmentation is different from conventional relational databases. We have proposed the vertical class fragmentation to reflect the characteristics of object-oriented databases. In this paper, we define the attribute partitioning algorithm and describe the results of implementation and comparison.