Top positive review
15 people found this helpful
A bit dated
on 16 February 2014
Back in the day, this was one of my most-referenced references.
If you work exclusively in relational data warehouses, such as Oracle RDBMS, then it's indispensable. It is still brilliant, and if you want to be a serious BI practitioner, you need to /know/ this text back to front. You need to know it *today*.
But times have moved on. I haven't built a presentation layer for seven years that was based on a /relational/ data warehouse. I still build 'em; it's just that they're only feeders for multidimensional hypercubes.
Back in 2002 I used Analysis Services to build a hypercube with 100 million sales and stock facts for a big retailer. It had a couple of seconds response time. Three years later we were writing sophisticated predictive analytics with multidimensional extensions (MDX).
There's no mention of this in the book. Solid on ETL, great on some industry-standard structures (but see stuff by Len Silverston et al), there's no question any aspiring BI person needs to know everything in here. But it's not enough.
I want to see this book updated to describe how to create a physical non-indexed DW as a feeder for a hypercube. I want to see the power and speed advantages of hypercubes described. I want to hear about the dramatic improvements in the presentation layer possible through the use of MDX. I want to hear about the security and /organisational/ issues involved in allowing reporting developers access to the presentation layer.
In short, as a new edition, this seems to fall a bit short. Still recommended, but . . .