An optimized scheme for vertical fragmentation, allocation and replication of a distributed database

Abstract

Distributed databases have become the most essential technology for recent business organizations. One of the main issues of the distributed database management system (DDBMS) is the distributed database design. The performance and efficiency of a distributed database design largely depend on the fragmentation of global relations. It also depends on the allocation and replication of those fragments to several sites of the network. Earlier approaches suggest vertical fragmentation solutions based on the frequency of the user queries. However, these approaches have the limitation of higher complexity and availability of the frequency of user queries at the initial stage of the database design, iterative binary partitioning in case of n-ary partitioning which will increase the complexity and predetermined threshold value that is used to enhance the results, as well as the problem of more computation. Consequently in this paper, we present a vertical fragmentation, allocation and replication scheme of a distributed database called (VFAR). The proposed scheme partitions the distributed database relations vertically at the initial stage of distributed database design, without the need for the frequency of user queries which are not available at this stage. It also allocates and replicates the resulted fragments to the sites of the distributed database, using manipulates and read operations of each site to each fragment. Experimental results showed that firstly; the proposed scheme reduces the overheads and the complexity of complicated computations of previous vertical partitioning approaches. Secondly, it solves the problem of the frequency of the user queries that are not available at the initial stage. Thirdly, it solves the problem of iterative binary partitioning in the case of n-ary partitioning by generating all fragments of one iteration. Finally, it gets the optimal results without the need to determine a threshold value that is used to enhance the results.