Revisiting arguments for a three layered data warehousing architecture in the context of the Hadoop platform

摘要

Data warehousing has been accepted in many enterprises to arrange historical data, regularly provide reports, assist decision making, analyze data and mine potentially valuable information. Its architecture can be divided into several layers from operated databases to presentation interfaces. The data all around the world is being created and growing explosively, if storing data or building a data warehouse via conventional tools or platforms may be time-consuming and exorbitantly expensive. This paper will discuss a three-layered data warehousing architecture in a big data platform, in which the HDFS (Hadoop Distributed File System) and the MapReduce mechanisms have been being leveraged to store and manipulate data respectively.