Abstract
The main focus of this master's thesis is to mapping star schema to HBase. And using MapReduce to find out an efficient way to implement the star-join query processing. Star-join is a core operation in the data warehouse. There have been a number of pa-pers about moving SQL to NoSQL and lots of approach to implement join operation by the MapReduce framework for large-scale fault-tolerant data analysis. But only a little related work study star-join query processing on HBase with MapReduce envi-ronments. In this thesis, we introduce an algorithm for processing star-join query on HBase using MapReduce. In our approach, a star-join query is split into many sub-queries. The fact table, which is split by column, sequentially join corresponding dimension tables with foreign key and then all the intermediate result and measure columns merge together to be the result. To further reduce the disk and network I/O, we adopt bloom filters pass through in each phase. The experimental results from the cluster of HBase and Hadoop environments show that our approach outperforms ex-isting approaches by cascading with two-way join.