October 24, 2001

Enterprise Foundation

Both Oracle and IBM offer an array of BI features, but business variables such as organizational fit and total cost of ownership are the final determining factors

By Tim Quinlan

There are times that define true greatness in competition. This scenario is rare and should be enjoyed to its fullest; it happens almost before we realize it, but when it's over, we miss it terribly.

This situation is occurring now in a battle of database virtuosos. IBM and Oracle have been exchanging blows since the early 1980s in the RDBMS market. Microsoft is the junior partner, having entered this battle later in the decade and competing on only a single platform. Microsoft is like a light-heavyweight champ that's moving up a weight class and looking good in the process. IBM and Oracle, however, are competing on every major platform and are both at the top of their games.

In this two-part series, I'll analyze IBM DB2 Universal Database (UDB) v. 7.2 and Oracle9i as strategic platforms for business-critical applications, with some information about SQL Server 2000 included as well. This particular installment concentrates on the respective vendor approaches toward business intelligence (BI); in the next one, I'll focus on e-business.

PHILOSOPHIES AND FUNDAMENTAL DIFFERENCES

At a high level, the IBM and Oracle offerings are similar. DB2 UDB v. 7.2 and Oracle9i both include improvements that address the same three major areas: data management, BI, and e-business. They both want to be the center of your database universe, but the implementation details differ: Oracle wants all your corporate data to reside in its DBMS, including nondatabase files that can use its Internet File System. IBM also has this broad goal, but its approach is associated with a federated architecture that enables access to heterogeneous data sources.

These products are fun to differentiate from a cost perspective. There are at least two highly subjective and contradictory total cost of ownership (TCO) reports available, each of which proves one product to have a significantly lower TCO than the other. (See Resources.) Interestingly, Oracle decided in mid-June to lower the price of its database (which remains higher than IBM's). Apparently, Oracle felt that customers were coming to the wrong conclusion about TCO. TCO is a complex issue, and only you have the intimate knowledge of your specific IT and business requirements necessary to determine it.

SYSTEM ARCHITECTURE SYNERGIES

Contrary to popular belief, database technology is far from becoming a commodity; those who claim that nothing has changed in database technology do not truly understand the revolutionary advances of the past few years. (For example, since 1995 we've moved from object-oriented DBMSs, optimizers, and row-level locking to implementing object-oriented features, access to most corporate data, Java, extract, transform, and load [ETL], and online analytic processing [OLAP] in the database.) These DBMSs are complicated and require a great deal of specialized expertise.

Oracle and IBM's latest offerings are vast. You will not be able to make your DBMS decision based on a feature set or a technical reason such as which vendor has a better partitioning scheme - both DBMSs do most things quite well. Rather, your final decision will be based on business-related reasons, such as the IT composition of your company, your systems architecture, the nature of your applications, and the mindshare in your company, industry, and perhaps even place of business. For example, your feature-by-feature comparison of databases may favor one over another, but if the pool of available talent in your company is strong in the other database, you will need to closely examine the benefits that will be gained in both the short and long term by using the technology that is more feature friendly.

Basically, the impact of the database choice on your systems architecture (and vice versa) is one of the most important factors in a buying decision. You will need to look at synergies that may exist among disparate technologies such as Web servers, application servers, message-oriented middleware, packaged enterprise applications, and tools. These synergies may quickly point you to a single vendor to deliver your database solutions.

If you're concerned about operating system support, both vendors currently support the leading Unix platforms as well as the IBM OS/390 mainframe and the Windows 95, 98, 2000, and NT environments.

THE BI TOUR

BI features in the databases fall into seven categories: data warehouse management, performance, OLAP, access to external data, ETL, data mining, and object/relational (O/R) features that enrich BI.

Data warehouse management. Both Oracle and IBM emphasize data warehouse management functionality, which is a critical issue because their products have become so complex.

DB2 Warehouse Manager and Query Patroller governs queries, analyzes costs, manages resources, and tracks usage. The Data Warehouse Center is a tool integration center providing ETL, quality assurance, and cleansing features via Trillium Software technology, as well as a schema modeler that passes metadata to the OLAP integration kit. This modeler manages authorization and defines processes and data flows graphically. Metadata is provided through the Information Catalog Manager that, for example, can show source and target data and provide details about data transformations.

With 9i, Oracle has improved its data warehouse management considerably. Data Resource Manager supports proactive workload management; you can create groups of users to better manage workload by limiting percentage of CPU, parallel degree, number of concurrent queries, and elapsed time exceeded. Memory and parallelism are allocated dynamically based on query requirements and resources available. By simplifying statistics gathering, the need to manage undo (rollback) segments - a data warehouse administrator's nightmare - has gone away. Integrated administration is available through Oracle Enterprise Manager (OEM).

Performance. DB2 has been a leader in performance optimization for more than a decade. Oracle introduced the cost-based optimizer with version 7 of its database, and with 9i, it has improved the optimizer to a point where it now considers more than I/O costs (adding CPU and memory) in its estimates.

Both databases implement sophisticated data partitioning that can enhance performance through query parallelism. DB2 UDB Extended Enterprise Edition (EEE) can scale to very large databases (VLDBs) thanks to a shared-nothing architecture. Historically, Oracle's competitive offering, the shared-disk-based Oracle Parallel Server (OPS), has not scaled as well as EEE. In 9i, OPS has been redubbed Real Application Clusters (RAC) and offers improved scalability, performance, and manageability, as well as support for NUMA and clustered servers. These changes should popularize parallel Oracle implementations, and in doing so, will present interesting competition for UDB EEE.

OLAP. Both databases have OLAP features built directly into the core database engine. They can both support relational OLAP (ROLAP) through star joins and analytic functions such as cube, rollup, grouping, and rank, as well as provide more complex capabilities such as moving averages and sums.

DB2 has a Data Warehouse Center with a warehouse schema modeler that can pass metadata to the OLAP integration kit and Star Schema Builder. Beyond ROLAP, IBM has a long-standing partnership with Hyperion Solutions Corp. to use Essbase as DB2's multidimensional OLAP (MOLAP) server.

Oracle includes metadata in the OLAP catalog that describes a multidimensional model and that you can manage with OEM. This architecture has two integrated data sources managed from the integrated catalog: the RDBMS and the Analytic workspace (formerly Oracle Express), which is a MOLAP database for predictive analytic functions. Application development is provided in two ways: through a Java-based API and BI Beans. You can use JDeveloper to develop BI Beans and use Oracle9iAS to deploy it. The Java-based API provides low-level access to the OLAP server and will more likely be used by third-party vendors to create custom Java apps and applets. BI Beans are a higher-level API that provide components such as cross tabs and charts that can be used by most corporate developers. Both are available with version 9.0.1.

Essbase has been regarded as the best-of-breed tool in this space for a number of years; Oracle's approach is more proprietary. Oracle is playing catch-up here. Essbase has stood on its own for a number of years as an OLAP server regardless of the RDBMS behind it, but it is hard to imagine companies using Express in cases where Oracle is not the primary RDBMS.

• Privacy
• Your California Privacy Rights
• Copyright © 2008 United Business Media LLC