ABCDatawarehouse Modeling and DesignA
is a “subject-oriented, integrated, non-volatile, time variant collection of datain support of management decisions [Inm].” The end-users of a data warehouse are usually business analysts, as distinct from field personnel or call takers.
: What do you think is the skill profile of the data warehouse end-user?
OperationalDecision supportData Content
Current valuesArchival, summarized, calculated data
Application by applicationSubject areas across enterprise
Nature of Data
DynamicStatic until refreshed
Data Structure &Format
Complex; suitable for operationalcomputationSimple; suitable for business analysis
HighModerate to low
Updated on a field-by-field basisAccessed and manipulated; no directupdate
Highly structured repetitive processingHighly unstructured analytical processing
Sub-second to 2-3 secondsSeconds to minutesSource: [STG]
: Do the descriptions under “Data structure & format” fit in with the skill profiles of therespective end-users?A
is a scaled down deployment of a data warehouse that contains data focusing on adepartmental user’s analytical requirements. For example, the Ohio-based Huntington Bank Corporation set up a data mart for its general ledger system, to get the ledger system's functionalinformation to the bank's financial analysts and budget coordinators quickly.
is the process of examining data for trends and patterns that might have evadedhuman analysis. For example, Shoko’s Sunday circulars contained coupons advertising healthand beauty aids, consumables, and household chemicals, which were are all located on the left-hand side of the stores. Shoko’s data mining exercise revealed that people who were coming into shop gravitated to the left-hand side of the store for the promotional items and were notnecessarily shopping the whole store. Consequently, it added apparel promotions to the Sundaycirculars.An
on-line Analytical Processing (OLAP)
application is intended to provide end-users an abilityto perform any business logic and statistical analysis that is relevant. This analysis must happenfast, i.e., it must deliver most responses to users within about five seconds, with the simplestanalyses taking no more than one second and very few taking more than 20 seconds.
are non-relational DBMS products that are specialized for use for the kinds of queries in data warehouses. This is in contrast to using specialized analysis toolsthat run on top of a traditional RDBMS.What is the ROI for a data warehouse? A recent study [Fis] of 45 major companies by theInternational Data Corporation found an
three-year return on investment in datawarehouse systems of 401%. More instructive is the very wide range of returns reported by thecompanies, from 16,000 percent to
1,857 percent. Moral: data warehousing is not a silver bullet; use with care!Multi-dimensional data structures can be implemented with multidimensional databases or extended RDBMSs. Relational databases can support this structure through specific databasedesigns (schema), such as "star-schema", intended for multi-dimensional analysis and highlyindexed or summarized designs. These structures are sometimes referred to as relational OLAP(ROLAP)-based structures.ABC Ver 1.10Page 2