Online analytical processing (OLAP) has evolved to aid users in reformulating flat, relational data files into multidimensional data stores allowing more optimal data retrieval and analysis. Essentially, OLAP creates a 'hypercube' of information allowing more complex analysis than a traditionalflat database. The most common OLAP analysis techniques are 'slice and dice' and 'drill down'
The term OLAP was first used by Dr. E.F. Codd who defined 18 features of OLAP. Criticism of these features centered around the fact that the 18 OLAP characteristics were extremely similar to those available in the product offered by Codd's sponsoring vendor. Over time, other researchers and vendors have expanded the 'rules' or features of OLAP to over 50. From this list, a generally accepted and understandable definition of OLAP has evolved: FASMI, or Fast Analysis of Shared Multidimensional Information. Below are brief descriptions of each of the FASMI characteristics.
FAST refers to OLAPs target goal of delivering most user responses in five seconds or less. Many responses actually take only one second and only the most complex requests generate responses taking longer than 20 seconds. As large amounts of data are involved, various techniques are employed to achieve speed. Various products, for example, use pre-calculations, specialized data storage, or specific hardware requirements for this purpose.
ANALYSIS refers to the system's ability to handle any relevant business or statistical analysis for a given user. Frequently used analysis techniques are 'slice and dice' and 'drill down'.
SHARED refers to the ability of the system to maintain system confidentiality and concurrent update locking if multiple write access is used. Many OLAP products are read-only but the more advanced ones recognize the need for write capability and are capable of handling updates from multiple users in a timely, secure manner. At a basic level, SHARED also refers to the ability of multiple users to access the same data concurrently without having to duplicate files or obtain IS assistance.
MULTIDIMENSIONALITY is the key feature of all OLAP products. OLAP systems must provide a multidimensional, conceptual data view and support multiple data hierarchies to be effective. As the key component, multidimensionality requires that data be organized in terms of the organization's actual business dimensions. For example, for a sales company, data may be organized along the dimensions of product, customer, time, and geography. At the intersection of all dimensions, cells are found containing the relevant data. Cells are easily located by a user seeking based on the terms of the position sought. Similarly to a flat spreadsheet, any cell in the multidimensional model can be calculated using other cells in the model. Often times, however, cells will contain no data. For example, a customer may not purchase the product in every time frame. This is referred to as sparsity. Many vendors have found creative ways to avoid wasting space by saving numerous empty cells; these systems, however, may be slower.
INFORMATION refers to all the data and calculated information required by the user. Data capacity across existing OLAP products varies greatly and is affected by many factors including data access method and level of data duplication. In short, to be an effective tool, OLAP must contain the data required by the user and offer effective analysis techniques to make this information meaningful to the user.
In achieving FASMI objectives, various techniques are used by different vendors. These include client-server architecture, time series analysis, object orientation, parallel processing, optimized data storage, and multi-threading.
www.access.digex.net/~grimes/olap/olap.collections.html