Abstract
Since the late ’80s and early ’90s, database technologies have evolved to a new level of applications: online analytical processing (OLAP), where executive management can make quick and effective strategic decisions based on knowledge in terms of queries against large amounts of stored data. Some OLAP systems are also regarded as decision support systems (DSSs) or executive information systems (EIS). The traditional, well-established online transactional processing (OLTP) systems such as relational database management systems (RDBMS) mainly deal with mission-critical daily transactions. Typically, there are a large number of short, simple queries such as lookups, insertions, and deletions. The main focus is transaction throughput, consistency, concurrency, and failure recovery issues. OLAP systems, on the other hand, are mainly analytical and informational. OLAP systems are usually closely coupled with data warehouses, which can contain very large data sets that may include historical data as well as data integrated from different departments and geographical locations. So the sizes of data warehouses are usually significantly larger than common OLTP systems. In addition, the workloads of OLAP are quite different from those of traditional transaction systems: The queries are unpredictable and much more complicated. For example, an OLAP query could be, “For each type of car and each manufacturer, list market share change in terms of car sales between the first quarter of 2005 and the first quarter of 2006.” The purpose of these queries is not for the daily operational maintenance of data; instead, it is for deeper knowledge from data used for decision support.
Key Terms in this Chapter
Roll-Up and Drill-Down: These are common OLAP operations. Roll-up allows us to look at coarser, “big picture” data by dropping one or more dimensions or climbing up along the dimension hierarchies. A drill-down operation is the opposite of roll-up for more detailed data.
Dimension Hierarchy: This is the granularity level structure of an attribute (called a dimension) that tracks a measure. For example, in time, the dimension we may have is a year-month-day hierarchy.
Data Cube: A data cube is a multidimensional data model to view the data stored in a data warehouse.
Relational OLAP (ROLAP): ROLAP is a type of OLAP server that mainly uses relational technologies. Other types of OLAP servers include MOLAP (multidimensional OLAP) and HOLAP (hybrid OLAP).
Online Analytical Processing (OLAP): OLAP is mainly for analytical and informational purposes rather than operational purposes. Through a friendly interface and efficient query evaluation, OLAP can help enterprise managers make informative, strategic decisions.
Data Warehousing: Data warehousing is a technology to extract, clean, integrate, and load data into a central warehouse database; it uses front tools such as OLAP and data mining to query the system for decision support.
Statistics Trees: A tree data structure used to evaluate OLAP queries efficiently. The levels represent dimensions and the branching links represent attribute values of that dimensions. A star pointer represents ALL values. Dwarf trees and QC-trees are other examples of tree structures for evaluating OLAP queries.