|
Online auction giant eBay is looking to capitalize on the expertise it has gained in running one of the world's largest data warehouses by offering business intelligence as an on-demand service.
Also See:
5 Members Join Petabyte Club
Join The CIOZone's Business Intelligence Group
By Mel Duvall
Online auction giant eBay is looking to capitalize on the expertise it has gained in running one of the world's largest data warehouses by offering analytics as a service.
The analytics as a service capability is first being offered to internal business units, but in time the company envisions being able to offer sophisticated business intelligence capabilities to outside parties, much the way Amazon.com and Google are pioneering cloud computing as a commercial offering.
EBay runs a 5-petabyte data warehouse to store and analyze data gained through running its online auction business. The data warehouse is currently used by some 5,000 analysts in more than 50 business units at the San Jose, Calif.—based company.
At the annual Teradata Partners conference in Las Vegas earlier this month, Oliver Ratzesberger, senior director of architecture and operations for eBay, shared how the company conceived, designed and implemented the new capability. When asked whether eBay was considering opening up its data warehouse to make business intelligence a commercial offering, he said that was being seriously considered.
"It's not something we're prepared to offer now, but in two years—yes, I think you may see us go there," Ratzesberger said.
"Our goal [with the data warehouse] is to be able to run at full capacity as much as possible. We want to extract every dollar we can out of our investment," he said.
EBay overhauled its data warehouse and analytics capabilities in 2002, moving from an Oracle-based system to a Teradata system. Ratzesberger said the move at the time was based on necessity—daily data loads were exceeding 24 hours and the system simply could not produce results in satisfactory time frames. The company uploads some 50 terabytes of new data each day.
While the Teradata upgrade helped overcome capacity limitations, Ratzesberger said it was still difficult for central IT to keep up with demands by business units for new data marts to perform analytics. A data mart is set of data pulled from operations or other sources and then used to perform analysis in a particular area, such as marketing. About two years ago, the idea was conceived to create a self-help service, where individual business units could use a Web-portal to load their own data into the Teradata warehouse. Once loaded they could then begin performing their own analysis using standard tools from such vendors as SAS, Microstrategy, Cognos or Excel.
Business units can also bring in outside data from third parties using the Web portal.
The end result is that instead of taking weeks for central IT to process a request, a business unit can upload their own data and quickly begin getting results, said Ratzesberger. "There literally are some business units who have an idea on Monday, and results by Wednesday," he said.
The idea of allowing business units to tap directly into the data warehouse in this manner did cause some uneasiness within the IT ranks. "Think about it—you have people bringing their own data in and messing around with your infrastructure," he said. As safeguards, the virtual datamarts are set up in sandboxes, what eBay calls PETs, short for prototype environments. Time limits of 90 days are set on the use of the PETs, preventing storage space from being indefinitely tied up and individual business units are limited to the number of PETs they can run.
If a prototype is deemed to be a success and of continuing value, the business unit can request that it be brought into full production. Since the parameters of the data mart are already known, the prototypes can be brought into production in a fraction of the time than if they had started from scratch.
Ratzesberger said business intelligence and analytics are vital to the current operations and the future of eBay. Analysts use the data warehouse to power queries into such areas as fraud detection, customer loyalty, user behavior, seller and buyer experience, and the efficiency of its own technology operations. "Every decision we make is based on analytics," he added. "It goes far beyond finance or marketing or the normal uses you might think of and into such difficult behavioral areas as what people are doing on the Web site."
Only registered users can write comments.
Please login or register. |
