Check this out too:
Overnight Data Marts
Information Week
October 27, 1997
Federal Express has spent years mastering the use of transaction-processing systems to track the millions of packages it handles each day. Now the shipper is taking the raw data generated by those big systems and building multiple data warehouses and data marts, in an ambitious effort to extend data-analysis capabilities to all corners of the company.
The starting point was a data warehouse FedEx deployed 18 months ago to support target marketing. That database, called the Information SuperHub, culls data on customers and their shipping habits from FedEx's billing system, then augments it with demographic and other data purchased from information brokers such as Dun & Bradstreet. The SuperHub, which has swelled to nearly a terabyte of data, won for FedEx one of six "best practice" awards this year from the Data Warehousing Institute, an educational organization in Bethesda, Md.
Now FedEx plans to deploy a half-dozen more data warehouses over the next several years. Each will serve as a repository for data in a different subject area-ranging from employee information to data gleaned from package scans-ensuring that the information available to all departments in the company is uniform and accurate. "The idea is to get everyone to go to a common source for each kind of warehouse information," says Cynthia Hubard Spangler, VP of corporate headquarters systems.
To maintain acceptable performance levels, FedEx will limit the number of users allowed to run queries against the warehouses. The warehouses will feed dozens of departmental data marts, which will be the data-analysis engines for most users in the company.
It's an architecture that will make it possible to extend decision-support capabilities to thousands of users. Today, only about 100 power users in the marketing department have access to the Information Superhub. In the future, says Hubard Spangler, "we want to extend business information to every employee that needs it. That could mean all employees in some fashion or another."
Impressive Scale
"What they're doing is very aggressive," says Richard Winter, a data-warehousing consultant in Boston. "I have not heard of another example of a company that is building multiple data warehouses on that scale and using them to feed dozens of data marts."
FedEx's existing Information SuperHub is based on an Oracle7 database running on a 46-node IBM SP2 server. The next warehouse-Hubard Spangler declines to say which subject area it will serve-is being built on an Informix Extended Parallel Server database and Sun Microsystems E 10000 server. FedEx has not standardized on one warehouse platform. "Sometimes volume is important. Sometimes it's speed," Hubard Spangler says. "We want to have more than one answer."
Likewise, FedEx has decided against using a single vendor's data mart package. FedEx does plan, however, to make wide use of Information Advantage's DecisionSuite desktop and server products, which support online analytical processing from Web browsers.
Done right, FedEx's data warehouse strategy could make a huge contribution to the company's overall business. "The payoff comes when data warehouses are being used in thousands of different actions by thousands of different people," says consultant Winter. "That's the formula for big-time success." |
