A Novel Vertical Fragmentation, Replication and Allocation Model in DDBSs

Hassan I. Abdalla (King Saud University, Saudi Arabia)

Ali A. Amer (King Saud University, Saudi Arabia)

Hassan Mathkour (King Saud University, Saudi Arabia)

Abstract: Modern database systems are commonly distributed, and data is kept at isolated locations (sites). The various sites are connected through communications links, which may be of low speed resulting in bottlenecks for data transfer between sites. Data replication is considered as one of the effective methods in dealing with such situations to achieve improved performance in distributed database systems (DDBSs). In this work, authors explore a new model for improving performance in distributed database environment by using a vertical fragmentation method along with a novel replication and allocation techniques. The solution procedure consists of a new vertical fragmentation model to fragment a relation and two phases of allocation of fragments to nodes. The paper discusses the tradeoffs between the different scenarios for finding an optimal way of deciding on attribute allocation to sites by evaluating performance based on the collected requirements. This model will significantly reduce communication cost and query response time in DDBSs.

Keywords: Vertical Fragmentation, allocation, clustering, distributed DBMS, frequency-matrix, heuristics, replication

Categories: H.2, H.2.8, H.3