A FRAGMENTATION TECHNIQUE IN DISTRUBUTED DATABASES & ITS IMPLEMENTATION

Abstract

The distributed database design is an optimization problem that includes various subproblems: data fragmentation, allocation and local optimization. Each of these can be solved in various ways. In context of this research we address fragmentation and allocation problems simultaneously for distributed databases management systems (DDBMS) for enhancing the process of server load balancing using Cisco Global Site Selector (GSS). Fragmentation is mainly of three types namely: horizontal, vertical and hybrid (grid) fragmentation. Although there is large work carried out on the design of data fragmentation but most of them are either horizontal or vertical. The core of this thesis defines a new type of grid fragmentation technique in the context of relational databases. Our proposal combines horizontal fragmentation (HF) on the basis of attribute locality precedence and cost model of vertical fragmentation (VF) to generate grid fragments. Both of these two techniques also address the problem of fragment allocation in distributed databases. As part of whole process we discuss and compare various other algorithms for generating candidate vertical and horizontal partition schemes. In the end, we compare our approach with one of the graphical grid fragmentation methods by implementing these techniques as part of Cisco GSS, a product for enhancing the DNS resolution process. Results shows that proposed grid fragmentation is a superior approach and it can solve fragmentation and allocation problem of relational database systems properly.