Parallel Real-Time OLAP On Cloud Platforms

Successful organizations increasingly rely on data analysis to develop new opportunities, guide business strategies and optimize resources. Online analytical processing (OLAP) systems are one of the most powerful technologies to provide the ability to interactively analyze multidimensional data from multiple perspectives. In this thesis we designed a new data structure, the PDCR-tree, that work on distributed systems providing low-latency transactions processing even for very complex queries. Using a PDCR-tree we demonstrate how to build a real-time OLAP system on a cloud based distributed platform called CR-OLAP. The CR-OLAP can be built using an m+1 machine scalable architecture so as the system load increases, the number of machines, m, can be increased to improve performance. Experiments show the system can process a query with 60% data coverage on a database with 80 million data tuples with a response time 0.3 seconds or less, well within the parameters of a real-time system.