Relational Data Access for Business Data Analytics

Data, information, and knowledge are dramatically increasing (Korth & Silberschatz, 1997; Naydenova & Kaloyanova, 2010). Although, Data Warehouses are a central access for business data since the 90s, new technologies have to be considered to achieve an efficient access and processing of these multidimensional data. Recently, new architectures have evolved that try to optimize data access for certain applications. Database systems (DBS) are pervasively used for all business domains. Therefore, DBS have to manage a huge amount of different requirements for heterogeneous application domains. New data management approaches are continuously developed, e.g., new trends are NoSQL-DBMSs (Chang et al., 2006; DeCandia et al., 2007), MapReduce (Dean & Ghemawat, 2008), Cloud Computing (Armbrust et al., 2009; Foster, Zhao, Raicu, & Lu, 2009; Buyya, Yeo, & Venugopal, 2008), to make the growing amount of data manageable for new application domains. However, these approaches are developed for specific applications and need a high degree of expert knowledge. From a technical point of view, there exist different opportunities to access, process, and analyze data in a more efficient way. On the one hand, the usage of hardware, due to decreasing cost is often a suitable way. On the other hand, this requires techniques that are developed or optimized for main memory usage in data warehousing. Another possibility is to use specialized storage models. Thus, it is possible to store data on all aggregation levels in multidimensional online analytical processing (MOLAP), e.g., the data cube, or only the most interesting data, e.g., iceberg cubes. Furthermore, the data access can be enhanced by considering the architecture. Since the 70s, relational database systems use row stores, that means, tuples (rows of a table) are stored sequentially. In contrast, column stores store data in such a way that attributes (columns of a table) are stored sequentially. This enables efficiency for data access in the domain of data warehousing due to a better access for aggregations. Another challenging optimization is the selection of a suitable index structure. Multi-dimensional analyses have to be supported. Dependent on domain, data, and application scenario different index structures can enhance data processing. In this chapter, we provide an overview of architectural decision. We focus on the storage architecture for relational database systems.