Univar, an industrial chemical company, had been on an extended acquisition trail going back to 1986 when it acquired McKesson Chemical. By 2007, when the company acquired CHEMCENTRAL Corp., Univar found itself operating a network of more than 170 distribution centers throughout North America, Europe, and China.
An application modernization effort begun in 2003 led Univar to migrate from VSAM to DB2 on System z. The organization also had a decade-old, Oracle-based data warehouse long overdue for modernization. The IT group wanted to modernize the data warehouse as part of a Business Intelligence (BI) effort led by Kevin Campbell, Univar’s manager of Application and Data Architecture. However, other more compelling business cases, namely acquisitions, trumped what was perceived as IT plumbing.
But BI is more than mere IT plumbing. It’s the way companies convert the tons of data they generate and collect into valuable business actions. By 2007, following its latest acquisition, Univar realized it needed to provide some kind of BI capabilities to its workers. Specifically, it needed analytical capabilities to understand its new Canadian acquisition. But a downturn in its global business ruled out Univar embarking on an enterprisewide strategic BI initiative.
With no long-term BI roadmap or any strategic BI initiative on the horizon, Campbell’s group decided to pursue a low-cost, skunk works, tactical BI project that leveraged its existing assets, mainly a System z9, DB2, a handful of specialty processors, and some basic reporting software.
BI on System z
BI is a natural on System z because that’s where most of the important production data live and where most of the processing happens. Mainframe reports might be considered the first BI, but when the PC emerged, with VisiCalc as a key application, BI quickly moved off the mainframe.
Today, BI remains mainly a distributed platform application. Spreadsheets, Graphical User Interface (GUI) query tools, Online Analytical Processing (OLAP), and multi-dimensional analytics fueled by data pulled from data marts and data warehouses constitute the bulk of BI. Transaction data travels through layers of middleware before being converted into meaningful BI.
With the advent of relational databases such as DB2, BI should have stayed on the mainframe, close to where the data was processed, stored, and managed. However, it wasn’t until the emergence of Service-Oriented Architecture (SOA) that the mainframe even began to regain a role in BI. SOA simplified the process, moving data directly out of the mainframe for use in analytics.
Enter IBM and Cognos
IBM became serious about BI when it acquired Cognos late in 2007. It didn’t get serious about BI on the System z, however, until the company introduced IBM Cognos 8 BI for Linux on System z early in 2009.