Article Preview
Top1. Introduction
In the present information age, data and information are vital not only for the survival of any corporate entity, but also for providing it with an edge over its competitors (Boisot and Canals, 2004; Rowley, 2007). In the context of today's global business and commerce, and innovative management paradigms, data and information have become the key currencies of corporations and governments worldwide. Secure and efficient data management, and fast information synthesis, have become all the more challenging as volumes of digital data are rapidly growing with time (Inmon, 2003). Database management systems (DBMS) were developed to efficiently manage and store the transactional data of various business processes of a department. Different departments of a corporation autonomously employed DBMSs that were purchased from different independent vendors. As a consequence, the data managed and stored in the departments were consistent and reliable for information synthesis only at the departmental level and not at the corporate level. This was so because, when data of one department was shared with other departments, the former seldom propagated its subsequent updates to those with whom it had shared its data. On the other hand, on account of incompatible DBMSs employed by different departments, it was never easy to efficiently propagate the updates. Such practice greatly compromised the consistency and reliability of data at the corporate level (Inman, 2003). Other factors like incompatible data formats, problems of homonyms and synonyms, heterogeneous operating systems and heterogeneous computer network protocols at departmental level, further aggravated the problem of data integrity at the corporate level. Information synthesis at corporate level from departmental databases became prohibitively expensive in terms of time, money and quality of corporate information; this resulted in an information crisis (Inmon, 2003; Psomas et al., 2002). To address such an information crisis, the data warehouse was developed whereupon data from various departments is collected, and then organized, based on business subjects. Such data is then transformed, cleaned and integrated to achieve consistent and reliable data. They are then loaded into the data warehouse. Every update of the data is time stamped and stored for historical analysis. Reliable corporate information can then be synthesized by processing the appropriate data from the data warehouse (Inmon, 2003). Thus, a data warehouse is the foundational database that provides reliable and consistent data at the corporate level to meet all the informational requirements of a corporation. In order to extract the necessary information from the data warehouse, it has to be analyzed from multiple dimensions. For this reason, a multidimensional model is used in data warehouse to store data (Gray et al., 1996; Shukla et al., 1998; Wremble and Koncilia, 2007). In a relational DBMS, data is stored using the star schema. It consists of many dimension tables connected to a single fact table. Generally, fact tables are massive tables containing millions of rows (Shukla et al., 1998; Shukla et al., 2000; Wremble and Koncilia, 2007). To support timely and effective decision making, data in the data warehouse is analyzed regularly to extract the necessary information; such data analysis, which requires response times of a few seconds, is called On-line Analytical Processing (OLAP) (Inmon, 2003). OLAP queries are not predefined; they are posed in an ad hoc manner by data analysts in response to the information acquired during analysis. To extract new information, analysts pose OLAP queries in an exploratory manner. With the rapid growth in the volume of data in the data warehouse, the response times of OLAP queries increase substantially due to the numerous expensive join operations between massive base tables of the data warehouse; in fact it takes hours and days to process such OLAP queries (Gupta et al., 1997). Such delayed response times of OLAP queries, not only render the data analysis process unsatisfactory, but also makes the obtained information stale and obsolete and, often, not useful for making any real time strategic business decision (Inmon, 2003; Psomas et al., 2002). With the view to improve the OLAP query response time, many techniques like indexing, query optimizer, query evalutation and materialized views have been proposed in literature (Harinarayan et al., 1996). The aim of the paper is to use materialized views to minimize the OLAP query response time.