Relation between datawarehousing and OLAP

I am currently looking to get some insight into the BI-world (in the context
of SQL Server), and I find the distinction between datawarehousing and OLAP
somewhat vague.
My say would be that OLAP could be a part of a data warehouse, but it is not
clear to me where the between the two is to be drawn:
What are your thoughts?
Tor B?dshaug
tor.badshaug(AT)bekk.noOLAP is the analytical portion those overlays you warehouse. It allows
for extremely fast & easy access to your warehouse for both IT and more
importantly your NON-IT personnel.
For the most part business analysts are not required to be proficient
in warehouse programming languages like SQL. So, OLAP data cubes give
them what they need without wasting time (money) and it aids in
insuring consistency & integrity of the data.|||The flow of data is typically as follows:
Production Database Servers -> (Staging DB/DTS/SSIS) -> Relational Data
Warehouse -> Analysis Services (OLAP) -> End User Reporting Application
The relational data warehouse is used to store data in a de-normalised
format (dimensional model), optimised for query. The relational data
warehouse can be used by reporting applications, but you will probably do
most of your reporting from the Analysis Services database.
Analysis Services databases are fast, intelligent and use an efficient,
compressed storage structure. Many reporting tools have been developed that
make use of OLAP databases - The metadata exposed by analysis services
enables them to understand the data. This allows you to buy a 3rd party
reporting tool such as Cognos, Business Objects, Pro Clarity and set them up
to connect to your analysis services database with minimal configuration.
These tools offer great flexibility and should allow end users (with
appropriate training) to answer 99% of their own queries. Most these tools
have simple to use, drag and drop interfaces.
A simple example of intelligence - You can define a display format for a
measure so it displays with a ￡ (or $) sign and comma separators - client
tools can read this metadata and know how to display the information to the
user. This example is really only the tip of the ice berg - thing get more
interesting with calculated measures, drill throughs and drill actions.
Another example is that OLAP databases are aggregate aware - creating
aggregates is possible in a relational database, but difficult to manage and
typically involves hard coding applications to use aggregate tables (unless
you write your own aggregate navigator)
Analysis Services is a part of the data warehouse - you can have a data
warehouse without an Analysis Services/OLAP database. You can also have an
analysis services database without the relational data warehouse component,
although this approach is not recommended. Some queries are still easier to
answer from a relational database and the relational database provides a
level of insurance if there is a problem with your OLAP database.
Re-Building your Analysis Services database is a simple task if you have a
relational data warehouse and a cube definition. Also, your task of
developing the Analysis Services database is greatly simplified if your data
is structured in a star/snowflake schema in a relational data warehouse.
Hope this helps,
David
PS
The BOL tutorials for Analysis Services 2005 are quite good to get you up to
speed with the basics of SSAS 2005.
David Wiseman
MCSE, MCSA, MCDBA
http://www.wisesoft.co.uk
"Tor B?dshaug" wrote:

> I am currently looking to get some insight into the BI-world (in the conte
xt
> of SQL Server), and I find the distinction between datawarehousing and OLA
P
> somewhat vague.
> My say would be that OLAP could be a part of a data warehouse, but it is n
ot
> clear to me where the between the two is to be drawn:
> What are your thoughts?
> Tor B?dshaug
> tor.badshaug(AT)bekk.no