I have purchased a number of books and not seen much of help from the book sources. I, although a big fan of books, had better luck learning online on the subject. The books I have, and I have a lot of them, tend to cover the basics of the concepts and not the design. Perhaps this is because your design has to take into consideration what tools you use. For example if you use Business Objects tools, they don't support molap while Cognos does. Then you have to ask yourself do you need molap? Unless your dealing with very large fact tables I would argue you don't. If your using Star or Snowflake schema there are not a lot of "design considerations". The more dimensions you have the more granular your database the slower it is but the more flexible it is. Your design should account for a mix of the clients needs, the speed available to you, how much volume of data you have and the speed of your toolset. For example the MS tool Analysis Services vs say Cognos or Microstrategies is like comparing a snail to a jet aircraft. But if you don't have a large volume of data, the snail can cross the finish line before the jet aircraft is even warmed up. Simple suggest is design your star schema with as few dimensions as possible and as summarized of facts as you can get away with. This limits the number of rows in the resulting star or snowflake schema design. This is on of the main design considerations in Data warehousing and Star or Snowflake schema design.
There are about three people who were very important in the Data Warehouse world. The first and foremost was: Ralph Kimbal. I read him first but don't 100% follow his claims. He over uses star schema design. He comes from big IBM thinking, I prefer a more federated design for my needs. If I was on a mainframe, what he suggests totally makes sense, but I am not. If you agree with this, then you move into the writings of: William(Bill) Iman. There is a third and up an coming individual whose name escapes me right now be he is taking the teachings of Kimball, and Iman and improving on them a bit. Mixing the varied schools of thoughts along with third normal form from E.F. Codd to make a faster design thats a bit of hybrid of relational design combined with star or snowflake schema.
Hope this helps...... In short start reading from Kimball and Iman but don't follow either of them blindly. Their worlds may not the world your developing in. In the end use there teachings as ideas as tools or methodologies to do what your tasked with.
Data warehouse is a world, you can't get it from just one book. Multiple theories and ideas. If you want just one book start with one from Kimball.
|
