The new face of data warehousing: Page 2

(Page 2 of 3)

Evolving into knowledge management

John Ladley, senior program director with the META Group and VP of Knowledge InterSpace, made two presentations at the recent DCI Data Warehouse and Knowledge Management symposium held in Phoenix this past December. The title of one, "Life After the Data Warehouse: Exploiting the Information Cycle" conveys the gist of Ladley's perceptive message: second generation data warehouses are being designed to be part of a dynamic information supply chain (ISC), and the ISC will ultimately morph into something we can refer to as knowledge management (KM).

There are plenty of definitions of KM (Ladley cited the META Group's, which says that KM is "...a discipline that promotes an integrated and collaborative approach to the process of information asset creation, capture, organization, access, and use. Information assets include databases, documents, and, most importantly, the uncaptured, tacit expertise and experience resident in individual workers..."), but Ladley offered a useful taxonomy. There are five main routes by which data warehousing develops into KM:

Process Route: Create improved cycle times, lower costs, and improve quality.
Product Route: Create, package, and market unique, higher margin products.
Enabler Route: Foster employee growth and empowerment.
Intellectual Capital Route: Prolong product leadership and embed knowledge into products and services.
Competitive Weapon Route: Capture competitive intelligence and differentiate yourself from the competition.

Ladley and other speakers at the KM track urged attendees to focus on just a "piece" of KM (one of the "routes"), or the one that offers the highest return. He could have cited his boss, Aaron Zornes, executive vice president of META Group, "The challenge today is to take business data from mind to market."
--Karen Watterson

Migration Architect 2.0 performs a combination of automated discovery and interactive analysis in order to generate a normalized data warehouse with atomic data. "It got us 95% there, right out of the box," according to ACSC's Nordstrom. Although he wishes Migration Architect would do a better job "preparing" the data for analysis (getting VSAM files into Migration Architect for analysis), Nordstrom was impressed with how easy it was to do column profiling (down columns), dependency profiling (across rows), and redundancy profiling (across tables and/or data sources). Evoke plans to add some utilities to help the "up front" loading of mainframe flat files in its next version, due mid-year 1999. It will probably also add Teradata and IMS to the list of databases currently supported (Oracle, Informix, Sybase, and DB2). In that version, customers will be able to use a traditional relational database for Migration Architect's datastore. Migration Architect 2.0 currently uses a proprietary data structure that even Evoke's vice president of engineering, Jack Olson, admits is "messy."

The secret behind Migration Architect is that it does one thing very well: data profiling and mapping. It doesn't do the data movement--users need other tools to do that. ACSC chose Passport to move its data. What Migration Architect does is help users figure out and "normalize" raw source data. This goes beyond what data cleansing software does--parsing, standardizing, and reformatting data, and often augmenting it with external demographic or financial data. Data cleansing is used by different people to mean different things, but it usually only refers to quality and consolidation issues.

Bruised and bloodied no more

Just over a year ago, in November 1997, the auto club began--working half days only--using Migration Architect to profile its data. "I wanted a tool that would help us transform the source data into third normal form (3NF)," says Norstrom. Less than six weeks later, Nordstrom and his team generated the desired sales report. "On New Year's Eve, we delivered the 1997 sales report," he recalls. "We were the knights in shining armor."

ACSC's new data warehouse now houses atomic data, while its datamarts are for sales and customer service. Nordstrom is proud of the auto club's hub and spoke architecture. And rightly so, as the industry-accepted architecture for "good" enterprise data warehouses and dependent datamarts, this type of architecture works well. Migration Architect profiles and maps ACSC's data while Carleton's Passport populates both its enterprise data warehouse, residing in DB2 on the mainframe, and its sales and POS datamarts. These datamarts, which contain records of member "transactions" such as ordering maps and "TripTiks," are in Oracle 7.3 star schema databases running under HP-UX. (Prepared by AAA for its members, TripTiks are customized route maps for a specific trip, which is broken up into a series of mini maps that are shaped like big tickets, hence the name.)

Now, with not one, but two datamarts under his belt, Nordstrom is ready to tackle what will be the biggest challenge of all--the insurance datamart. A few years ago, auto club staffers estimated it would take between three and four staff years to design a normalized database from data residing in four VSAM files and 58 different record types. Using Migration Architect, Nordstrom hopes to accomplish the task in just 12 months--with three major deliverables along the way at 120-day intervals.

Like ACSC, Leslie Cone, project manager at the U.S. Department of the Interior Bureau of Land Management, also discovered Evoke by accident, thanks to a consultant's recommendation. About a year ago, it was clear that the BLM's extremely ambitious long-term reengineering project, the Automated Land and Minerals Record System (ALMRS), was bogged down in a quagmire of horrendously complex data conversions.

ALMRS, which had been initiated almost a decade ago and contracted to Computer Sciences Corp. in 1993, was to have been completed in 1997. Last summer, when it was obvious that ALMRS was taking longer than expected, BLM put the project on hold in favor of a "must complete by March 1999" rehosting project.





Page 2 of 3

Previous Page
1 2 3
Next Page





0 Comments (click to add your comment)
Comment and Contribute

 


(Maximum characters: 1200). You have characters left.