Relating to OLAPOLAP and ROLAP are a continuum, not competitorsby Joy Mundy Continued from Page 1 But woe betides the user who wants total sales for an arbitrary period such as January 3 through March 12, 2002, for which no predefined hierarchy exists. Part of the blame belongs to the client query tools, some of which won't even let you formulate the query except by bringing back daily data for January, February, and March. If your business users really need to do this a lot, you should include this functionality in your product review matrix. Great AdvantagesI just discussed some things that are really easy in the pure relational world but are difficult for some OLAP servers to handle. But OLAP is brimming with advantages compared to relational systems. Here's my list of OLAP's advantages:
As a simple example, consider inventory balances: The inventory balance for January and February is certainly not the sum of inventories in January and February. You can train users not to sum inventory balances over time, but will they always remember? Will all users use the same aggregation method, such as ending or average balance? OLAP systems can handle this problem transparently. An OLAP system lets you have server-defined calculations of great complexity. SQL's limitations as an analytic language were outlined in a previous column. SQL is not an analytic or report-writing language: You need an analysis server to support statistics, data mining algorithms, or even simple rule-based business calculations such as allocations and distributions. The OLAP server acts as a friendly interface to the data cube, letting users consume server-defined analytics without worrying about how and where they are defined and computed. Server-defined, high-performance queries and calculations can be performed over multiple fact tables or cubes. Combining data from multiple fact tables is a difficult problem in the pure relational world, but can be made easy and intuitive in certain OLAP servers. Calculations can be defined once and used many times. The more calculations you can define on a central server, the more flexibility your users have in accessing the data. Even a simple slice-and-dice tool can use complex analytics previously defined on the OLAP server. This capability is not generally found in relational environments. And of course, power users can define complex calculations on the server so all users benefit. OLAP is just plain fun. Most of the time. Designed for AnalysisRecent trends in the OLAP market are toward lower cost, improved performance and scalability, increased functionality in the core analytic space, and extensions to neighboring spaces such as data mining. These trends will continue over the next few years as the major database vendors bet more heavily on OLAP servers and more tightly integrate those servers with other data management and analytic software. OLAP servers present dimensional data in an intuitive way, enabling a broad range of analytic users to slice and dice data to uncover interesting information. OLAP is a sibling of dimensional models in the relational database, with intelligence about relationships and calculations defined on the server, that enable faster query performance and more interesting analytics from a broad range of query tools. You shouldn't think of an OLAP server as a competitor to a relational data warehouse, but rather an extension. Let the relational database do what it does best: provide storage and management. Don't torture yourself forcing the RDBMS — and its clunky query language SQL — do something they were not designed for: analysis. Guest columnist Joy Mundy [joy@microsoft.com] evangelizes data warehousing and business intelligence best practices for Microsoft SQL Server.
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| |||||||||||||||||||||||||||||||
