Enterprise FoundationBoth 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 Continued from Page 1 Accessing external data. The list here is impressive. Both products support Web clickstream data, data replication, and MQSeries, although Oracle has its own Advanced Queuing product. Oracle accesses flat files outside of the database as "external tables" that are read only and can be used for transformations. IBM has implemented access to multimedia files through Data Links that put the file under control of the database server and replicate them in a synchronized manner. DB2 supports distributed queries to SQL Server, Sybase Adaptive Server, and Oracle through DB2 RelationalConnect (from DataJoiner) built into the database. Distributed query access is now available from SQL procedures. Metadata interchange is provided through Common Warehouse Metamodel XML objects and ERwin to IBM products such as Data Warehouse Center and Information Catalog Manager. Oracle offers gateways to access external data and a distributed query option is now available as well. IBM and Oracle are lobbying Microsoft developers heavily. Both databases support Object Linking and Embedding (OLE) DB. Oracle9i supports Microsoft Transaction Services and Internet Information Service (IIS) access, among other things. One of the biggest surprises here is that DB2 now supports stored procedures written in Visual Basic. DB2 UDB Data Warehouse Manager Connectors can be used to access new application sources such as the i2 TradeMatrix BPI or SAP R/3 by clicking and dropping an SAP object into the process modeler. DB2 UDB also excels at accessing heterogeneous data sources using wrappers to access other DBMSs as a federated database, and MQSeries with SQL statements using functions. These features will allow your business to integrate and analyze disparate data in a manner that will appear to be more seamless. ETL. The fact that ETL operations now exist inside the DBMS is a large change in the way we think about databases. This process is the most difficult one in BI. It forces us to merge data that does not match and that often comes from noncomplementary data models. Although designers like to focus on the more visible and rewarding star schemas, ETL design and processes tend to be overlooked and underestimated. In fact, performing ETL and merging data across systems can feel like trying to combine two completely different jigsaw puzzles into a single picture. DB2 UDB has Visual Warehouse integrated into it; you can use more than 100 advanced transformations to clean and transform data. Oracle's ETL functionality is a set of features more than an integrated toolset. It provides Warehouse Builder, of Carleton legacy, to design and manage ETL processes and the transformation engine for end-to-end ETL processing. Other features - such as a multitable insert, a new Merge statement to insert and update in a single pass, and the ability to access external tables and table functions that take a set of rows as input and produce a set of rows as output - assist in the process. Change Data Capture allows Oracle data changes to be processed without using the entire set of source rows. Data mining. Data mining is a critical BI component that helps you understand data patterns and predict future behavior, and it's built into each database. UDB's Intelligent Miner for Text provides scoring to give mining deployment capability and is implemented as an extender. Predictive Model Markup Language (PMML) is used to express analytic models, and the model logic is accessed through a standard SQL API. The scoring process can be created as an SQL view and can operate on Oracle databases. The Oracle9i Data Mining Suite and data mining algorithms, which originate from its Darwin acquisition, include customer churn, customer profiling, response modeling, and profitability modeling. The algorithms include Transactional Naive Bayes for supervised learning to determine a future expected value or class, and predictive association rules for unsupervised learning to reveal clusters and associations in data. You can use scoring to predict outcomes, and data mining algorithms are embedded in a Java-based API. 9i Personalization can dynamically serve up recommendations to end users based on data mining. For example, customers visiting a Web page as repeat visitors could have a series of products presented to them that might be of interest based on mining data from previous Web site visits. O/R and content management. Both UDB and Oracle have extensive O/R features. Both allow proprietary indexing to be developed that can support BI. DB2 UDB has Extenders for spatial or text information, as well as a Life Sciences data connect for biotech data sources. Oracle has Data Cartridges (similar to Extenders) and multimedia offerings for image, audio, and video through Oracle InterMedia. DIFFERENTIATING FACTORSBoth companies have integrated the major BI capabilities into their DBMSs. IBM has taken a best-of-breed, federated approach (examples are Essbase and ERwin integration); Oracle, although it supports open standards, is more inclined to implement all data content in Oracle databases and make customers manage them with Oracle tools (such as JDeveloper and Oracle9iAS). It's impossible to discuss DB2 UDB without touching on IBM's acquisition of Informix. IBM's official stand at the time of this writing is that Informix products (such as Red Brick Decision Server) will be supported for the foreseeable future. This claim is entirely believable. What is difficult to determine is how long this period will last, and what will happen next. Both databases offer a range of BI features. However, beware of brand-new features in Oracle9i, as it can take a few releases to shake out the bugs. For IBM's part, some UDB features operate on a subset of platforms. For example, only UDB for Windows supports Visual Basic stored procedures. Furthermore, not all features I've mentioned come free with each database. Your challenge will be to determine the true TCO for your organization, including all extra licensing charges and salary and consulting expenses. You will then need to determine how each database fits into your corporate architecture. For example, a shop that primarily runs Oracle applications, an Oracle RDBMS for OLTP access, and Oracle Application Server should use the Oracle BI solutions, whereas a shop that does not have a large investment in Oracle and runs WebSphere, AIX, MQSeries, CORBA, and a mixture of diverse databases and data sources as its OLTP RDBMS should strongly consider UDB. Obviously, most shops are somewhere in the middle, but the combination of TCO and architectural fit will be your final deciding factors. Tim Quinlan [tquinlan@tlqconsulting.com] is a database consultant based in Toronto. He has more than 20 years of experience designing and implementing OLTP and data warehousing databases. Thanks to Bill Wong of IBM, Greg Martell of Oracle, John Eng of Microsoft, and Grant Zolkavich of North Technology Partners for their assistance. SQL SERVER RISINGMicrosoft's bundling strategy may have landed the company in hot water in the past, but it has been responsible for catapulting its DBMS to a new level and has pushed competitors to do the same. Starting with SQL Server 7, Microsoft integrated many BI tools into the RDBMS as a standard offering. These features included OLAP (including ROLAP, MOLAP, and HOLAP) and data mining offerings through Analysis Services, ETL through Data Transformation Services, and English language query, shared metadata, and easy-to-perform warehouse management. Rather than treating this bundling of features as an unfair competitive practice, IBM and Oracle followed in a similar fashion with their more recent offerings. Of particular importance is Microsoft's leadership in integration, pricing, and ease-of-use, which have pushed IBM and Oracle to try to match them. According to John Eng, lead product manager for SQL Server BI, Microsoft's strategy is to make data warehousing "easier." Thus, the company's approach was to build a comprehensive platform in-house that would be fast-to-market, easy to use, and easy to deploy. Building its products in-house differentiates Microsoft from other vendors that have purchased technologies and integrated them into their databases. Another differentiating factor is SQL Server's bent toward proprietary Microsoft platforms such as COM, .Net, and ActiveX Data Objects. Oracle and IBM are placing more emphasis on open solutions such as CORBA and Java. In the future, you can expect to see a continued focus on delivering a database that is easier to program and facilitates application delivery. Microsoft is also pushing for a large increase in terabyte databases. This will be interesting to watch: Microsoft has successfully proved that its database is integrated and easy to manage. But it must now prove to skeptics that SQL Server and Windows have climbed the scalability and availability ladder to a point matching that of Unix and mainframes. RESOURCESBuyers Guide to Database Servers Based on Cost of Ownership and Effectiveness, by Analyst Firm Input: www.input.com/buyers_guide/yntco/yntco_main.cfm DB2 UDB vs. Oracle8i: Total Cost of Ownership, D.H. Brown and Associates: www-4.ibm.com/software/data/oracledb2.pdf IBM DB2 UDB: www-4.ibm.com/software/data/db2/udb Microsoft SQL Server: www.microsoft.com/sql/default.asp Oracle9i: www.oracle.com/ip/index.html Related Article at IntelligentEnterprise.com: "Data Mining for the People,"May 7, 2001: www.intelligententerprise.com/010130/products1_1.jhtml
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| ||||||||||||||||||||||||||||||||
