sql server data warehouse: Hi,

There are many points concerning the BI world that are very confused for me.

-> What is a relational DB ?
A relational DB is for example Oracle, SQLServer or mySQL isn\'t it ? So on
a relation DB, we can model an entity/association DB (especially for OLTP)
or a Star/snow flake schema for a datawarehouse ? Is that makes senses ?

-> If I\'m right, a datawarehouse requires an isolated database (in order to
not affect production systems) that contains a star schema database (because
a star schema provide response in less time), doesn\'t it ?

-> Concerning OLAP, I always read that relational DBs are not designed for
multi dimensional analyses. So is that means that neither E/A schema nor
star/snow flake schema are adapted for OLAP ? That also means OLAP tools
must build their own OLAP DB before providing analyse functions ?

Are there some mistakes in this post ?

Thx a lot

NightFox

You're right, the answer is No

Analysis Services, Microsofts OLAP system, contains the data in its own
datastore, not in the SQL server store.

[quoted text, click to view]

[quoted text, click to view]
Hi NightFox,
there are lots of books around on BI now and I've posted a lot of materials
on my web site www.peternolan.com for beginners. There is even a beginners
page with links to the books I have read and recommend. (I don't recommend
books I have not read.)

An RDBMS? eg SQL Server, DB2, Oracle, Sybase ASE etc. The number1 book in
this area is 'Introduction to relational databases' or similar by CJ Date.

There is no truth in the rumor that a DW must have a separate 'cube' based
database to perform olap processing. This hasn't been true since 1984 when
Metaphor Computer Systems (co-founded by Ralph Kimball) produced the first
ROLAP system. This was a system that could do OLAP processing on a
relational database. Today, ALL major database players have built in the
capability to perform OLAP in the relational database and hence it is called
ROLAP. However, designers of DWs can use ROLAP, or MOLAP which is a 'cube'
based database for OLAP and even 'H'ybrid which is call HOLAP.

The designer must choose whether to go for both a relational database and an
analytical cube based database for a specific implementation.....I have done
many DWs where we have had both and many where we had no 'cube' based
database. The cubes are ahead in speed and functionlaity for OLAP
processing. One of the main reasons being they do not have to be ANSI SQL
compliant so the vendors can put functions into them much more quickly than
they can RDBMSs. The cubes give blinding performance and ALL the good tools
talk to the cube products. For example, you would be very hard pushed to
find a query/analysis tool that did not talk to MS Analysis Services.....

The cube products and the database both have their places in the world of
DW/BI.

Best Regards

Peter