TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

IBM Announces Linux-powered Data Warehousing Bundles

If IBM’s new Linux-powered bundles look and sound like data warehousing appliances, are they, in fact, data warehousing appliances?

• By Stephen Swoyer
• August 31, 2005

Summer is traditionally a fallow period in the business intelligence space, but IBM Corp. has been very busy over the last few months.

In June, Big Blue announced a new packaged data warehousing bundle, the IBM Data Warehousing Balanced Configuration Unit (BCU), based on its pSeries Unix servers and TotalStorage hardware. Last month, IBM shifted into BI overdrive, terminating the development of its DB2 OLAP Server, nabbing customer data integration (CDI) specialist DWL, and introducing an open-source enterprise search technology, UIMA.

Big Blue had another trick up its sleeve, too: It expanded its BCU bundles to include two popular Linux distributions. But if IBM’s new Linux-friendly BCUs look like an appliance, sound like an appliance, and even smell like an appliance, are they, in fact, data warehousing appliances?

Not according to Big Blue, which positions its Data Warehousing BCU as a cross-platform offering, albeit one that includes proprietary servers, software, and storage. There’s some truth to this claim, too. Although IBM’s first BCU was an AIX-only offering, Big Blue now ships BCU packages based on Red Hat Linux and Novell Inc.’s SuSE Linux.

At this point, however, BCUs are only available on IBM pSeries hardware, with IBM TotalStorage arrays and DB2 database software elsewhere in the backend. Big Blue’s pSeries systems, for the record, are powered by IBM’s proprietary Power5 microprocessors. In this respect, the BCU concept sounds a lot like a data warehousing appliance from Netezza Inc. or DATAllegro Inc. In fact, IBM’s all-in-one BCU proposition eerily resembles that of Teradata, the NCR Corp. subsidiary that markets a line of proprietary (i.e., hardware, software, and storage) data warehousing systems.

This has encouraged some industry watchers to make the inevitable comparison between the two—or four. “Although IBM is not using the ‘appliance’ term, BCUs, with their preconfigured server, memory, storage, operating system, database, and associated utilities, could easily be seen as data warehouse appliances,” writes Mike Schiff, a senior analyst with consultancy Current Analysis. “As such, they provide additional credibility in support of the appliance concept, while also representing IBM’s response to competition not only from appliance specialists such as Netezza and DATAllegro, but also from established top-end enterprise-class data warehousing specialists such as Teradata.”

As far as Schiff is concerned, IBM’s BCU vision is a compelling one: As canned building blocks for creating data warehouses, BCUs should simplify many implementation and integration issues—at least for customers in the market for an IBM-centric data warehousing solution.

Underneath the covers, the BCU is based on IBM’s pSeries p5 Model 575 eight-way servers, IBM’s TotalStorage DS4500 storage array, and DB2 Data Warehouse Edition. It’s populated with 1.9 GHz Power5 processors—Power5 is, by all accounts, one of the fastest and most scalable processor architectures—and ships with 32 GB of RAM standard.

Industry veteran Schiff says the AIX BCU makes for a tempting package. “By combining and integrating the underlying disk, memory, CPU, and I/O components based on the experience of the IBM BI Best Practices Team in designing and configuring data warehouse implementations, IBM has optimized them for data warehouse environments,” he writes. “BCUs are designed for use as modular building blocks for creating data warehouses. Since DB2 UDB utilizes a ‘shared-nothing’ architecture, it should be relatively easy to combine BCUs in a highly scalable fashion.”