Reference architecture speeds data warehouse deployment.
How do you roll out a major enterprise application in a timely and cost-effective manner when simply deploying and testing the hardware for the new system can take months and add way too many dollars to the final project cost?
For Kim Ross, CIO at Nielsen Media Research Inc., the answer was to get his hardware vendor, Sun Microsystems Inc., to build and test the server and storage platforms on which Nielsen is now building a 20-terabyte data warehouse to hold and analyze information on television viewing habits.
The decision to useand, in fact, help developSuns iForce Enterprise Data Warehouse Reference Architecture will dramatically reduce the amount of time Nielsen would have had to spend setting up and testing a new data warehouse architecture, said Ross. The pretested and pre-tuned setup will also allow Nielsens data warehouse to be more efficient, reducing storage requirements by as much as 65 percent.
"The proof-of-concept demonstration put together in Suns iForce Ready Center gave us the ability to see for ourselves how effective the solution would be for our company," said Ross, in Dunedin, Fla. "Reference architectures are a real boon to IT managers who are considering solutions for a specific business problem. For us, it further validated that what we wanted to do would work."
Nielsen isnt alone in its desire for turnkey, pretested reference architectures. Research company Gartner Inc. estimates 60 percent of enterprises will find their IT architecture is obsolete because of accelerating business and technology change and that many will turn to such solutions to retain competitive advantage.
Nielsen Media, which provides the Nielsen television ratings and is a subsidiary of VNU Inc., of Haarlem, Netherlands, began last year designing the 20-terabyte data warehouse that will serve as the underpinning of a new decision support application. The system will allow customerstypically marketing and advertising companies in the broadcasting and cable industriesto query and perform analyses on television viewer data online.
Ross, who was already using Sun hardware and Sybase Inc. products, wanted to upgrade from Sybases Adaptive Server Enterprise database to the Sybase Adaptive Server IQ Multiplex analytical database to accommodate the increased performance needs of the new data warehouse. But, as Ross and his IT team began to design the system, they realized they did not have the hardware in-house to test certain aspects of the database. Ross, for example, wanted to test a feature in Adaptive Server IQ Multiplex that allows multiple production copies of the database to share a single storage resource.
Sun and Sybase in October built for Ross a proof-of-concept architecture in a Sun iForce Ready Center in Menlo Park, Calif., to demonstrate Ross proposed configuration and provide him with a blueprint of how he could build an enterprise-class data warehouse using their products. The iForce Enterprise Data Warehouse Reference Architecture that Sun eventually assembled runs on Sun midrange servers with Sun StorEdge 9960 systems and Sybases Adaptive Server IQ Multiplex analytical database. It is designed to handle more than 25 terabytes of raw data input. The servers and storage devices can be customized or changed based on a customers needs.
While the demonstration gave Ross a chance to see how the infrastructure would work at Nielsen, it did not provide applications to run on top of the reference architecture. Thats because, as currently offered, the iForce Enterprise Data Warehouse Reference Architecture includes only the underlying server and storage elements, not reference database schema designs, decision support or other tools. Nor is security included. Sun executives said that because applications are unique to each customers business requirement, the company provides customization through its professional services division and integrator partners.
"What you have is database storage and a database engine, but whats missing is an application to use your data," Ross said. "But it gets at the problem that they say theyre getting at, which is the essential way of how to build a database."
Ross said Nielsen engineers are building two application layers on top of the data warehouse. One will serve as the application layer customers use to access the database and will include query applications. The second is a statistical processing layer unique to the business rules Nielsen employees use to compile television ratings. Both layers will be written using Java.
The entire data warehouse will be housed on Nielsens corporate production LAN and secured in a data center. Nielsen plans to release new analytical products to its customers using the data warehouse as early as years end, something Ross said would have been impossible had he built everything himself.
"It takes time to build such a beast," Ross said. "When youre in the business of providing information to other businesses, your ratings are based on your ability to meet your customers needs."
As a senior writer for eWEEK Labs, Anne writes articles pertaining to IT professionals and the best practices for technology implementation. Anne covers the deployment issues and the business drivers related to technologies including databases, wireless, security and network operating systems. Anne joined eWeek in 1999 as a writer for eWeek's eBiz Strategies section before moving over to Labs in 2001. Prior to eWeek, she covered business and technology at the San Jose Mercury News and at the Contra Costa Times.