The Boulder BI Brain Trust

 

Richard Hackathorn: December 2008 Archives

Teradata Ponders Next Wave for Active Data Warehousing

| | TrackBacks (1)
TD logo.pngWe had a delightful and free-wheeling discussion with Stephen Brobst, CTO at Teradata. The topic is: Operational BI and Strategic BI on One Platform: Marketing or Reality? Stephen shared an early consulting engagement, 13 years ago, that exemplified an Operational BI (OBI) situation for him. It was a shipping distribution network of managing the flow of trucks from source to destination. The issue is the efficient shuffling of contents among trucks at key break points. This business problem is fairly generic across several industries; the solution is complex. For Stephen, this is the first instance of a valid business justification for extending the data warehouse from strategical analysis into operational analysis.

SB pic2.jpgDefinition of Terms: We kicked around the definition and boundaries of OBI. Claudia remarked, "Traditional BI stores the data and then analyzes it. OBI analyzes the data and then may store the data." Complex event processing (CEP) is analyzing on the fly...data in motion versus data at rest. One approach that I offered is to define OBI in terms of critical points in decision processes where multiple alternatives have low switching costs. Which alternative results in the greatest value to the enterprise? I think that the business need depends on rethinking the business process model. IOW you may need to reengineer the business to leverage OBI. Stephen hates the term "real-time BI" because it set the wrong expectations. Rick noted that we should move at the "speed of business" rather than in real-time.

Workload Management: Stephen characterized OBI workloads as: availability (7x24), performance (sub-minute), data freshest ("more") ...implying that the major cost was keeping the data fresh. The assumption is that queries for OBI are known. OBI requires an embedded a "learning mechanism" and workflow management that incorporates this learning, such as refreshing/rescoring of the models.

Physical Data Model
: We lapsed into a brief wandering through Inmon-Kimbell issues and then sketched several architectures to support OBI -- separate ODS-EDW and single EDW embedded with OBI.

Synthesis: Operational BI is not BI. It really is about redesigning and enhancing business processes by incorporating the historical context. The purpose of this context is to determine business value in a specific decision situation.

This was a great discussion! The above captures only a small fraction of the morning. I wish future BBBT sessions were as thought-provoking.  

 

 
   

 

This page is a archive of recent entries written by Richard Hackathorn in December 2008.

Richard Hackathorn: October 2008 is the previous archive.

Richard Hackathorn: January 2009 is the next archive.

Find recent content on the main index or look in the archives to find all content.