Online Analytical Processing Architecture


Conceptually, OLAP architecture consists of three functional components : presentation services, OLAP services, and database services (Figure 12.3).

Figure 12.3. Functional Components of OLAP Architecture

graphics/12fig03.gif

Presentation Services

OLAP is supposed to provide the link everyone has been looking for between data and the business. Yet most organizations still seem to be data rich and information poor because the real world of business exists in bits and pieces, not in bits and bytes. Information is data that can be analyzed , synthesized , and used in a valid business context. The people who need this data are knowledge workers, business analysts, business managers, and business executives, not technicians. Therefore, the data needs to be presented in a format that enables the business people to develop proposals, decide how many widgets to buy, define investment levels, and set hiring targets.

Presentation services have to be easy to use, and ease of use has to be determined by the business people, not by the information technology (IT) staff. For example, business people want an intuitive graphical user interface (GUI) and the ability to work with familiar business terms. Therefore, an easy-to-use OLAP tool should hide the underlying structure of the data and hide the processes that run behind the scenes. Furthermore, ease of use should be expressed in quantifiable terms.

  • How much time is needed to learn the OLAP tool?

  • How fast can a person perform his or her analysis tasks ?

  • Do business people like using the OLAP tool?

  • Does the OLAP tool have all of the required functionality?

  • Can the OLAP tool integrate with other desktop tools, such as Microsoft Excel?

Presentation services have to be flexible and adaptable because different business people have different preferences and different skill sets. For example, some business people like tabular reports ; others like graphs and charts . Some business people have no computer skills at all, some are more advanced, and some are experts. The menus , icons, and functions should be configured depending on the skill set profile, and they may have to be reconfigured over time. When the beginners start to become experts, they no longer like the cute messages that were originally provided to them for encouragement. Experts expect better performance and faster responses, and in order to provide that, there should be less clutter on their screens. An ideal OLAP tool should be able to adjust to all these different levels of preferences and skill sets and should be able to provide different levels of presentation.

OLAP Services

An OLAP tool should provide a wide range of services. It should be able to support simple querying with just a few dimensions, and at the same time, it should be able to support powerful querying with many dimensions. In addition, an OLAP tool should be able to integrate all the analytical processing requirements of "What happened ?" with those of "Why did this happen?" Querying capabilities (from very simple to complex), reporting capabilities (from very basic to sophisticated), and multidimensional analysis and presentation of the results are some of the OLAP services that help turn data into useful information.

Querying, reporting, and analyzing are interrelated, interactive, and iterative. For example, the results of a query might appear in the form of a table, chart, or graph, presented in several dimensions. While studying these query results, a business analyst may think of a new question, which may lead to a new query. He or she may then want to have the results of the new query printed out as a report. Therefore, OLAP tools should have integrated querying, reporting, and analyzing services. A person should not have to log off the querying tool to get into a different reporting tool and then log off the reporting tool to get into an analysis tool. Querying, reporting, and analyzing should be a seamless transition performed by the tool, not by the person.

In order to leverage these OLAP services, we need to change the way we develop applications and the way we present information. BI applications, which emphasize quick delivery of functionality, ease of use, and affordable desktop hardware and software, should be the vehicles for IT to provide OLAP capabilities to more business people in the organization.

Database Services

OLAP architecture supports two types of databases, conventional relational databases (e.g., DB2, Oracle), which are accessible with ROLAP tools, and proprietary multidimensional databases, which are supplied with MOLAP tools.

  • ROLAP tools can access any of the major relational DBMSs as long as the underlying application database design is multidimensional, such as star schemas (facts and denormalized dimensions), snowflake schemas (facts and normalized dimensions), and hybrid schemas (combination of normalized and denormalized dimensions). Depending on the DBMS, the database designer would use common physical design techniques such as:

    - Selecting the most efficient indexing schema for the underlying DBMS product in order to improve performance

    - Partitioning the database into smaller, manageable partitions to improve performance and to facilitate database maintenance activities

    - Clustering the data and physically co-locating related tables

    - Determining the most appropriate data and index placements

  • MOLAP tools are designed to access their own proprietary databases, which are special data structures (e.g., Essbase), and to perform their OLAP operations on these data structures. MOLAP tools implement their functionality in a variety of different ways.

    - Some products store data in arrays or cubes and therefore have different data preparation requirements.

    - Some products require prebuilding dimensions in the staging area before loading them; others build the dimensions from the data at load time.

    - Some products provide an application programming interface (API); others do not.

    - Some products offer "turnkey" applications with multidimensional servers and substantial OLAP functionality.

    - Most products have their own proprietary access methods and front ends.



Business Intelligence Roadmap
Business Intelligence Roadmap: The Complete Project Lifecycle for Decision-Support Applications
ISBN: 0201784203
EAN: 2147483647
Year: 2003
Pages: 202

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net