Beyond simple analysis: Doing more with data warehousing
- By Joab Jackson
- Jun 25, 2007
Although many people feel reusing failover databases is a good idea, some see the limitations of this approach.
'For every good question you answer, it leads to another good question,' said Bill Cooper, vice president at data warehouse system provider Teradata, a division of NCR.
Cooper warns that relying on a transactional database for analysis needs might only get you so far when you need to dig into the data. Organizations that execute reports against failover databases might want to consider bringing in historical data as well. A warehouse can hold data from the past few years ' data that might not be worth keeping in a primary database.
By bringing in additional database sources, organizations can gain greater
insight into long-term trends happening within the data. These trends might not be identifiable by looking at the material over a shorter time or comparing fewer factors. Technically speaking, data warehouse software can do 64-way joins, whereas most transactional databases can do six, eight or, at the most, 12 joins, Cooper said.
What this means is that, with some planning, you could examine the relationships among many more factors than the eight or 12 you are tracking in your database.
The Air Force, for instance, might want to know how often a particular engine experienced mechanical difficulties in the past six months. The number may be useful for predicting failures, but additional factors may provide even more insight. How many of these failures occurred in the desert? How many occurred in cold weather? What kind of maintenance was done before the failures? How many of these failures occurred in the past 10 years? The more factors you add, the better picture you will be able to paint, Cooper said.
'If you have an enterprise data warehouse in your architecture, when you do an inquiry you can do it not only against current data but the historical data as well, and that will get you a better answer,' he said.
Cooper also said more agencies are going in the direction of 'moving business intelligence from strategic to operational' concerns. This means going beyond compiling reports that compare how performance metrics stack up against some quantified mission of the agency ' and moving toward analyzing the operational aspects of the agency itself. Operational aspects may be something like how many aircraft are ready to fly and how much inventory would be needed to keep them running.
And again, data warehousing can provide a more suitable environment for keeping track of these factors, Cooper said.
Rob Ashe, chief executive officer at business intelligence software provider Cognos, has remarked on a similar trend. 'We're seeing a lot of interest in government on performance management,' he said. 'It used to be about reporting, but now it's about improving response time, or improving troop readiness.'
At one time, reports were generated that would, in a roundabout way, influence leaders' plans to better carry out organizational missions; now, metrics can be more tightly coupled to directing operations themselves.