Aggregation tool cuts wait for BI data at publisher
- 23 August, 2006 15:09
Until a few months ago, users at publisher Simon & Schuster had to wait as long as a half-hour for results from business intelligence queries about sales figures.
The New York-based publishing house was facing a problem that is becoming more prevalent as companies open up access to increasing numbers of business users: balancing the growing demand for data access with the strain put on BI tools and databases. In response, it installed a data aggregation tool to speed up data access.
The 50 users of Business Objects SA's enterprise reporting tools at Simon & Schuster were demanding ever-more-detailed reports from the company's growing data warehouse, said Stuart Mowat, vice president of information systems and technology. Because of the prolonged query response times that resulted, users were increasingly asking business analysts to run the reports.
To tackle the problem, Simon & Schuster began to implement a data aggregation tool from HyperRoll Inc.six months ago. Mowat said the tool integrates with the Business Objects software and processes query requests from it. Using data compression and aggregation algorithms to precompute a fraction of aggregated values, HyperRoll has improved query performance at the publisher in several areas. For example, response times to sales force queries have been reduced from 16 minutes to one second, and those for regional sales queries have decreased from 11 minutes to one second.
"The speed of the reports coming back is obviously the most important advantage," said Mowat.
Next month, the company will migrate from an Oracle Corp. database to Microsoft's SQL Server 2005, and it plans to use HyperRoll with SQL Server. In addition, Simon & Schuster plans to eventually expand the use of the data aggregation tool to all new areas in which it plans to analyze aggregated data, Mowat said.
In May, the company began using HyperRoll with the sales analysis system that it built internally using Microsoft Visual Basic and .Net. That system performs aggregated analysis on combined retailer and point-of-sale data.
"[It] was not usable before because of response times," Mowat said. "We were getting response times of a minute and a half, which for an online system was unacceptable." With HyperRoll, response times decreased to five seconds, he said.
The company is exploring the possibility of using HyperRoll to integrate historical information in its data warehouse with data from transactional systems to provide users with more up-to-date BI reports.
"A lot of our [workforce] is becoming less satisfied with a data warehouse that shows data updated as of the close of business [the day before]," Mowat said. "They are looking more and more for as close to real time as we can get it."
Bill Hostmann, an analyst at Gartner, said companies that roll out BI enterprisewide often encounter performance problems. Often, they constrain use rather than adjust their technology to accommodate the challenges, he said.
"When you go and tell the users to stop hitting the system," Hostmann said, "you are basically telling them to stop asking the questions about the information you need to run the business."