New Variants of Hash-Division Algorithm for Tolerant and Stratified Division

Works done in the context of the relational division for DBMS led to several approaches. Among which, the Hash-Division algorithm proved its superiority compared to the other approaches in the most of the cases. Nowadays, current trends of division are been oriented towards flexible queries and those involving preferences. However, the emphasis was always on proposing new operators which provide more flexibility and tolerance than the classical division operator. The performance aspect has not been adequately addressed. The proposed approaches in the literature suffer from a lack of performance, especially in a large volume of data. In this paper, we attempt to address this problem. Our idea consists in exploiting the advantages offered by the classical Hash-Division algorithm to propose new variants tailored for the flexible context. We paid a special attention to the improvement of some extended tolerant operators. Furthermore, we introduce a parallel implementation of our proposed techniques. Experimental results show the efficiency of our proposition. We obtained a very satisfactory improvement in processing time (the gain exceeds a ratio of 20 in the majority of cases) in both sequential and parallel implementation.