| System i suits Arcadia with WebFOCUS and DB2 | |
| 14 November 2007 Arcadia has improved business intelligence (BI) capabilities through utilising WebFOCUS and DB2 on its System i 570, according to a report from IBM.
A 2002 takeover of Arcadia Group brought a new management style and greater focus on strategic planning, leading to a surge in the use of data warehousing and BI tools. Arcadia needed to deliver more detailed BI more rapidly to growing numbers of users, while maintaining low infrastructure and operational costs. The data warehousing platform needed to be both stable and highly scalable, and offer continual performance enhancements.
The solution? Arcadia runs its 2.25Tb data warehouse on the i570 with the latest-generation POWER5+ processors. Data management is handled by DB2 for i5/OS, an integral component of i5/OS, and WebFOCUS is used for creating and delivering BI reports. Integration of DB2 database into i5/OS helps simplify data management and reduce costs, and the query performance enhancements in DB2 enable Arcadia to maintain a six-hour window for overnight batch jobs while data volumes have doubled.
“The System i platform has kept ahead of our rapid growth, and the continual improvements in DB2 optimisation on the platform mean we’re getting better performance even as data volumes and user numbers rise,” says Matthew Evans, development manager, Arcadia.
“As the number of users grew, and as their reporting requirements became more diverse, we wanted to enable greater flexibility in development,” says Evans. “We were not confident our existing reporting tool would scale adequately, and in development terms it was something of a black box: a change in one area might trigger an inexplicable degradation in performance.”
Arcadia selected WebFOCUS from Information Builders as its new reporting and BI tool, running the software on System i. “WebFOCUS is very flexible -- it’s almost a language for our development team,” says Evans. “The ability to tailor reporting to a very fine degree will become increasingly important as the data set grows.”
The deployment of WebFOCUS in a logical partition alongside the data warehouse on i570 enables very fine-grained performance tuning, and takes advantage of highly-secure virtual Ethernet connections to speed data delivery and keep traffic off the LAN.
“One of our tables now stands at 700Gb, with 650m records,” says Evans. “WebFOCUS with DB2 data warehouse on i570 produces extremely good response times for users running reports on this very large table.
“The agreement between IBM and Information Builders to integrate WebFOCUS into the standard System i software stack as DB2 Web Query for System i vindicates our decision to invest in the software, and is likely to enable further optimisation of System i as a platform for data warehousing.”
When Arcadia introduced data warehousing, the company ran it on a Unix platform. Evans says: “There was significant effort involved in getting the hardware, OS and database working together reliably. Consequently the platform lacked the stability we needed, especially when compared to the systems we were running on System i.”
Arcadia had moved its data warehouse to System i after a successful test migration at the DB2 Centre of Competency in Rochester. “System i was clearly more than capable of delivering the performance and scalability required for a data warehouse. Ever since, System i has kept ahead of our rapid growth, and continual improvements in DB2 optimisation on the platform mean we’re getting better performance even as data volumes and user numbers rise.”
The data warehouse holds all sales and inventory-related data across the group, and enables reporting down to the level of individual stores. Over the past three years, the number of registered users of the data warehouse has grown from 150 to over 600, and the volume of data doubled to 2.25Tb.
The Rochester Centre of Competency runs a cut-down version of Arcadia’s data warehouse, enabling the DB2 developers to test new optimisation techniques on a real-life customer system. “We’re currently working with Rochester to tweak the performance of our data warehouse for i5/OS V5R4,” says Evans. “We’re conservative about new OS releases, because we cannot risk increasing the time taken to load the overnight batch. The batch window is around six hours at peak times, and we need to complete processing before the first users come online.”
Frank Booty | |
