The Challenge Of Making Business Intelligence And Analytics Work
Posted by
Mary Shacklett
October 27, 2011
A major European metro subway records and stores video camera images in a database, and uses an analytics software package to monitor tracks and tunnels for any abnormal visual pattern—such as a package that seems to be sitting too long in a particular area that may need to be investigated.
A package delivery company uses analytics software in conjunction with GPS and the digital tracking of trucks to evaluate when trucks are taking too long to run their routes and why.
A major supply chain sourcer analyzes available shipment options for least cost and fastest delivery on a 24/7 basis, and makes at-the-moment, real-time logistics decisions.
The list goes on and crosses every industry. For IT, it means making infrastructure and technology performance decisions that can accommodate the business’s demands for the same real-time performance from its analytics and decision support systems as it has demanded in the past from its transaction processing.
How do you get there?
Servers--If you are going to infuse the business with real-time analytics, you can’t get there by reprovisioning servers that were purchased and deployed for processing large volumes of transactions because analytics processing is a different function. One reason is that transaction processing servers read data in rows instead of in columns. As Sybase director of analytics Don Lahl pointed out, this is like "entering a UPS truck in the Indy 500." Analytics servers do read data by rows, but most of the time they drill down for information by looking at columns of data that contain data attributes such as "sales" to point you to sales data. Reading by column instead of by row is faster and more expedient for retrieving query information because you’re bypassing data topics that don’t matter to your query.
Page: « Previous Page | 1 | 2 | 3 | Next Page »
