META Group Sees "Hub-And-Spoke" Architecture As Solution To Data
Mart-Oriented Warehouse Design Through 2000                          10.07.97
ACTION ITEMS                                                           D S *

Stamford, Conn. -- META Group, Inc. announced that recent client interviews and market research have revealed that, although solving the performance problems found in large centralized data warehouses, data marts have created new and costly problems for IT.

As an alternative to high-volume querying of monolithic databases, data marts have proliferated, leaving IT with the complex task of keeping shared data current and timely across all subsets. According to META Group, the multiplicity of data marts is creating interface management complexity, data movement and coordination problems, and an associated cost explosion.

"In effect, we are regressing to the old 'islands of data' problem that a central data warehouse was supposed to solve," said Doug Lynn, senior research analyst in META Group's Open Computing & Server Strategies service. "What we're seeing is that the data mart model is failing to give users timely access to universally current data that can be shared as needed.''

"Hub-And-Spoke" Architecture

META Group believes that by 2001, delivery of query services will be provided by a "hub and spoke" architecture that will properly match query requirements with database management system (DBMS) capabilities, contain cost, and provide much needed flexibility to meet evolving informational requirements.

To implement this design, META Group is advising its clients to engineer an information mover infrastructure (IMinf) that loads transaction data from operational systems into the hub, a central atomic data store. The hub will then consolidate, prepare and deliver data down each spoke as determined by each line of business. On each spoke, there will reside a DBMS that loads the new data from the hub, performs view maintenance, and builds or rebuilds indices, as necessary.

"We believe that this model will provide essential data sharing across data models and DBMS technologies without the manual use of elaborate and expensive products," said Lynn. "The enterprise will realize currency of data across all 'spoke' query processing facilities, and the IT organization will be able to contain the cost and complexity of data mart development and maintenance."

Headquartered in Stamford, CT, META Group, Inc. is an independent market assessment company providing research, analysis, and advisory services regarding developments and trends in information technology. META Group currently serves more than 1,300 clients worldwide. Additional information about META Group is available on the World Wide Web at http://www.metagroup.com