A Fork in theBy Mel Duvall | Posted 2007-05-14 Email Print
Re-Thinking HR: What Every CIO Needs to Know About Tomorrow's Workforce
Wesco International didn't want to pony up big bucks for traditional enterprise resource planning system. How a new, quicker data warehouse helped the $5.3 billion company anchor its business.
Road">A FORK IN THE ROAD
In the end, Conte and Van Oss decided that Wesco didn't need a new ERP system. It would have meant a wholesale replacement of existing systems for gathering data, transacting business, and analysis and reporting. What Wesco really needed was a faster way to get at the data already being sent in to headquarters via WesNet.
The company decided to replace its Informix data warehouse, which had served well since it was installed in 1993 but was long past its prime, with an Oracle data warehouse a version of the Oracle database optimized for analysis of large volumes of data. An enterprise licensing agreement was signed with Oracle in February 1999, and construction of the new data warehouse began.
That's when events took an unexpected turn.
Wesco's NCR account representative got wind of the company's Oracle migration plans and asked Van Oss whether he had considered using an NCR Teradata system instead. (NCR announced plans in January to spin off Teradata as a standalone company.) Conte and Van Oss didn't want to consider changing tack at this stage construction of the Oracle data warehouse had already begun and, just as important, financial approval for the project had already been granted by Wesco's executive board. But out of courtesy to NCR's 20-year relationship with the company, they agreed to a benchmarking exercise.
"Privately, Steve and I said, 'The chances of their being able to make a significant difference over Oracle are slim, so let's just humor them,'" Conte recalls.
Conte handed Teradata a sample set of data required to produce their most challenging report the month-end sales analysis report that previously took 80 hours to complete. In tests using the Oracle data warehouse, Conte and his team had been able to get the analysis report down to 28 hours. "We were so pleased with that, we actually called the report Kickass," he says. With further tweaks, Conte thought they might be able to get the report processing down to 24 hours a single-day turnaround but he wasn't hopeful to make much progress beyond that.
"I said to Teradata , 'If you guys really want to impress us, here's where you can do it,'" Conte remembers.
After about seven weeks of testing, Teradata contacted Conte to say it could get the report processing down to three or four hours. "I said, 'Hold on, you can't be doing it right.' Then they said, 'We think we can get it down to two hours,'" Conte says.
Conte and Van Oss looked at the Teradata results and determined that the results were correct. In fact, in the months ahead they would be able to lower the processing time to 75 minutes.
After a hard deliberation, Conte and Van Oss called the team together and put a halt to the Oracle implementation. "I had already made the decision that if the results of the benchmarking test were just close, we would continue down the path to finish the Oracle data warehouse ," Conte says. "But these results weren't close they were miles apart. We realized we had to stop."
Baseline contacted Oracle to comment on Wesco's decision to implement Teradata and to discuss differences between the two database systems, but in an e-mail, Oracle representatives chose not to comment.
Wesco decided to continue implementing Oracle for some functions, particularly transactional data such as pricing, electronic data interchange (EDI) and the company's e-commerce environment. The Oracle system also feeds information back into Teradata. But Teradata would now serve as the storage hub for sales analysis, accounts receivable and payable, supplier summaries and customer master records.
The reason for this two-pronged approach is basically the same as the explanation for why Teradata was faster than Oracle at producing the sales analysis report the data warehouses are built differently. Although both are built on relational database management system (RDBMS) technology, in which data is organized around related tables (rows and columns) of data, the design of the Teradata RDBMS has always revolved around fast analysis and retrieval of data. Oracle grew up around online transaction processing (OLTP) applications, in which the most important thing is to record transactions such as purchases and payments quickly and reliably. As a general-purpose database, Oracle can also be tuned and configured to support more analytical applications such as data warehousing.
However, because Teradata databases are almost exclusively focused on data analysis, the database software is highly optimized for that singular purpose. For instance, Teradata incorporates a technology known as massively parallel processing, in which database lookups are broken into smaller sub-tasks that are assigned to different processors on a multi-processor server; this technology, along with other factors, allows the database to respond to queries or produce reports very quickly.
In Wesco's case, the data being collected from its branches and analyzed fit more easily into the Teradata system. While it might have been possible to restructure the data for faster reporting from an Oracle database, that would have required a major retrofit of the WesNet system or perhaps some additional data integration and translation technology, according to Conte. The Teradata system was more expensive than a comparable Oracle system $5 million compared to about half that amount but Conte says the difference in processing speed was worth the higher price tag. And it was certainly a lot less than the ERP system the company had priced.
In December 2000, Wesco took delivery of a 1.3-terabyte Teradata Warehouse. Initially, the system ran parallel to the existing Informix system, but as the Teradata system proved itself, the Informix setup was phased out. In 2002, Wesco took delivery of a newer 1.8-terabyte data warehouse, and reassigned the initial Teradata system to application development.
One of the key technology decisions was choosing a tool for presenting information from the data warehouse and for conducting business intelligence queries. Initially, Wesco deployed software from Ottawa vendor Cognos, but eventually selected the WebFocus suite from New York-based Information Builders. Conte says both tools performed well in fact, some power users at Wesco still use Cognos but Wesco was able to secure a more cost-effective enterprise license from Information Builders.
The combination of the Teradata data warehouse and the WebFocus tools gave Wesco the two things that had long eluded the company: Closer to real-time access to data from field operations, and a way of drilling down into the data.
Conte's team tweaked WesNet at the branch level to push inventory updates to head office several times a day. That capability had always been possible, but it hadn't made sense to do it more often because the data couldn't be processed quickly enough.
Using the data warehouse as the backbone, Conte's team has since constructed a number of applications more often associated with ERP suites. Each morning, for example, when Wesco's salespeople log in to their e-mail accounts, they receive a report with the previous day's billings, and can drill down to see individual customer or supplier accounts. The e-mail has built-in links to WebFocus, which in turn taps into the Teradata data warehouse. Teradata has crunched the information from the nightly data feeds received overnight via WesNet.
Conte admits the de facto ERP system he has built doesn't have all of the functionality of a packaged system. Some applications, like contract approvals, come out of the box with most ERP suites. "But there are also a lot of things that come with ERP systems that we don't need," he says. For example, WesNet already provides Wesco with a sales and inventory system, something that would be duplicated by a full ERP installation.