Using Dimensional Hierarchy Analysis in Data Warehouse Fragmentation Process

Abstract

     The data warehouse becomes extremely large in size and its implementation normally uses very expensive platforms, typically based on high-end servers or high-performance cluster. Therefore, emerged the need to develop a distributed data warehouse to meet requirement specific to a departmental or restricted community of users.

     The transition from the using of a central data warehouse to the distributed data warehouse needs to fragmenting  the central data warehouse into several smaller warehouses (or data marts) physically distributed across the nodes in computer network. There are a lot of algorithms used in the process of fragmentation such as genetic algorithm, affinity-driven algorithm, etc. In this research we will use the principle of hierarchy located in the data warehouse as the basis of the fragmentation process, we proposed framework includes the representation of hierarchical property in the form of inverted tree with several levels depending on the existing data, so we can conduct searches on this tree using searching algorithms to get to the best level that could be adopted in the fragmentation process. In this framework a greedy search algorithm will be used because of its advantages to support the process of selecting the appropriate level that will use in fragmentation process.

     The use of distributed data warehouse as small data warehouses (data marts) after the use of fragmentation process was the first attempt to solve the problems of space, performance, and cost of special hardware