informa
/
News

How to Put the BI in 64 Bits?

The more an analytic tool can hold in virtual memory, the larger its data sets, such as OLAP cubes, can be -- and the faster it can perform.

The more an analytic tool can hold in virtual memory, the larger its data sets, such as OLAP cubes, can be — and the faster it can perform. Extract, transform and load (ETL) operations can bypass bandwidth limitations if they can cache large sets of data in virtual memory. A database infrastructure with ample addressing space can better handle compute-intensive activities like concurrent and complex queries. Optimization and decision models can also perform better. In other words, business intelligence (BI) is a fertile field for 64-bit computing. So why hasn't much cropped up?

Trendspotting

Performance management software once existed almost purely in the financial domain. But now it's increasingly finding a place in departments associated with operations, lines of business and management. Those departments include sales, marketing, customer service, operations and human resources.

Last year, only 30% of the customers of Applix, maker of the TM1 performance management suite, said they were using TM1 for nonfinancial applications. The 2005 poll shows that 53% now do.

Although the finance department usually has a greater affinity for management by numbers than most other departments, business results depend greatly on what happens on the operational level. More businesses seem willing now to exert the extra effort to quantify and track performance in less culturally prepared areas.

The explanation might have something to do with risk aversion on the part of vendors and buyers. For buyers, switching to 64-bit operating systems means accepting the small subset of drivers and security applications available for such environments.

Regardless, BI vendors have been slow to release 64-bit applications particularly to the new price/performance leaders, AMD Opteron and Intel Xeon EM64T. Part of the problem is deciding which 64-bit architecture to support, because each one requires different code compiles and reprogramming. Also there is the question of which operating system (OS) to support on the various architectures. The choice isn't trivial, because each OS has different 64-bit capabilities with regard to maximum memory, virtual memory space, dual core CPUs, virtual systems, and other 64-bit performance components. Finally, 64-bit BI applications may not yet have a large enough buyer base. A representative of BI vendor MicroStrategy recently admitted in an e-mail disclosed to Intelligent Enterprise that its 64-bit applications seem to be available ahead of demand, for example.

All hope is not lost, however. Two software technologies related to BI are leading the charge to 64 bits: compute-intensive analytics and, especially, databases. Databases were ported to 64-bit environments like HP Tru64, HP/PA-RISC and Sparc64 well before x86 hardware and OSs were available. So it's not surprising they're leading x86 64-bit conversions as well. Oracle and IBM have 64-bit Linux versions and Microsoft has recently catapulted into third place in the TPC benchmarks with its 64-bit version of SQL Server 2000 on Windows Server 2003 x64.

The hardware is also becoming more accessible, in terms of both availability and price. 64-bit x86 processors from both AMD and Intel are now making 64-bit computing practical for a much broader range of servers and workstations--and at well under $50,000 for 5GHz (total dual core) CPU, 5GB memory and 5TB disk space.

True, BI applications converted to 64-bit x86 architectures are still hard to find. Even SAS, with its great reputation among the technical elite, has so far limited its 64-bit versions to Itanium and PA-RISC. But SQL Server, IBM DB2 and Oracle 10g all have significant BI capabilities in their databases. Also, the MicroStrategy 8 BI platform has been ported to Red Hat Linux on x86 Xeon processors. And a few small companies such as QlikTech are promising 64-bit versions.

Because demand for 64-bit BI is bound to increase eventually simply because of the tidal wave of data coming from such sources as RFID deployments, business activity monitoring, financial monitoring and compliance monitoring, you have to wonder whether the database vendors will leap ahead of the pure-play BI vendors in the market. Database systems — engineered to be more scalable, secure and reliable than BI platforms — are naturally going to be stronger as organizations work with the larger data sets on 64-bit platforms. In other words, 64-bit computing could introduce a sea change in the BI industry every bit as dramatic as the upcoming entry of Microsoft into the market.

— Jacques Surveyer

Get Smart

Knowledge work is sometimes applied to increase business productivity, but rarely is the productivity of knowledge workers themselves reliably assessed. That's why it's so exciting that business process reengineering (BPR) guru Thomas H. Davenport has shifted his focus from the high-volume, easily observed and improved activities of BPR over to the much more nebulous task of improving knowledge workers' performance.

In Thinking for a Living (Harvard Business School Press, 2005), Davenport explains why the Frederick Taylor "efficiency expert" approach — although it revolutionized industrial age productivity and led to the successes of BPR — will most likely harm knowledge workers' performance. Yet, exempting knowledge workers from all manner of performance analysis isn't a good alternative, even though it's the one most commonly used. The (rare) studies done on white collar workers show them to be "grotesquely unproductive," in fellow guru Peter Drucker's words. What to do?

Davenport provides a truly useful framework for analyzing knowledge workers and methods for improving their work. He manages to combine theory with enough practical details to help you apply it soundly and to reassure you the theory itself is sound — all in a short, pleasant read.

— Jeanette Burriesci

Editor's Choice
Salvatore Salamone, Managing Editor, Network Computing
Mary E. Shacklett, President of Transworld Data
Joao-Pierre S. Ruth, Senior Writer
Richard Pallardy, Freelance Writer