Erkonen and Geiver said they knew after five months that the installation of the second array was more of a band-aid than a solution. Throwing more hardware at the problem wasn't going to make it go away, they reasoned at the timePremier Bankcard needed a new infrastructure if it was going to continue to gain market share.
"There was a fundamental shift in what the business was asking IT to do as analytics began to play a very large role in growing the business," Erkonen said. "We quickly identified that we'd need to rethink everything to support rapid growth for a sustained period of time."
Geiver and Erkonen decided that the only way to sustain Premier Bankcard's high rates of data growth was to deploy a centralized, unified business intelligence platform to support the 70 employees who have access to the data warehouse. In 2004, they developed a strategy that took advantage of 64-bit computing technologies to ensure the maintenance of a highly available and centralized data warehouse, speedy access to information in the data warehouse, improved application performance, and up-to-date data loads.
In June 2004, working with HP consultants, Premier Bankcard's IT staff began to overhaul its infrastructure and installed an HP Integrity rx8620-32 Server running Microsoft's 64-bit SQL Server 2000 for mission-critical business intelligence. An HP Integrity rx5670 Server was brought in to run SAS Enterprise Miner r8, while an HP Integrity rx4640-8 Server was deployed to run Microsoft Analysis Services. Premier Bankcard also decided to move to the 64-bit version of Windows Server 2003 in hopes of increasing performance and throughput.
Erkonen and Geiver took advantage of an existing HP Enterprise Virtual Array, adding 25.5TB to serve as the cornerstone of the company's SAN and provide centralized data access.
With a single data warehouse and a unified set of data, multiple employees can now simultaneously load files and query the data warehouse without contention. The queries that used to take as long as 22 hours now take roughly 45 minutes. A profitability query (a query that determines the profit derived from a particular customer) that used to take 13 minutes now takes as little as 36 seconds, Geiver said.
"One of our goals has always been to determine the profitability of individual credit card accounts in our portfolio, but that meant digging through millions of accounts and transactions," Geiver said. "We now have the data modeling, mining and analytical tools we need to [efficiently] determine profitability."
Erkonen said the time required to load data into the warehouse has decreased by 30 percent and that performance has improved by as much as 100 percent. Analysts now spend from 80 to 90 percent of their time analyzing data, up from 50 percent.
Next Page: Growth spurt.