Horizontal Partitioning of Multimedia Databases Using Hierarchical Agglomerative Clustering

Horizontal partitioning is a database design technique widely used in relational databases in order to achieve query optimization. Recently, this technique has been applied in multimedia databases to improve query execution cost in these databases. Nevertheless, current algorithms are based on affinity between predicates to obtain an horizontal partitioning scheme (HPS). Affinity measures how a pair of predicates is accessed by the queries ("togetherness"). The main disadvantage of this measure is that it only involves two predicates, and hence does not show the "togetherness" of more than two predicates. In this paper we propose an horizontal partitioning method for multimedia databases which is based on a hierarchical agglomerative clustering algorithm. The main advantage of our method is that it does not use affinity to create the HPS. We present experimental results to clarify the soundness of the proposed method.