Designing cube

[Former Member]

Hi Friends..

The below question asked by many project managers or team leads..but i couldnt explain in detail...can you help me to answer to this partiular question as iam new to BW..i said like collecting Characteristics and key figuers and assigning to dimensional tables..then the interviewer didnt satisfy what i said..he needs total explanation about modelling..how do you start first from the scratch..to cube and build query on that..

From the scratch how do u design a cube and how do you a get query?

[Former Member · Accepted solution]

Hi Raja !

First of all, don't forget to reward the answers given to your previous post (and for the future too !)...it's a way to say thanks and to assure gratitude from our BW experts...

if during an interview your answer is "designing a cube means collecting characteristics and key figuers and assigning to dimensional tables"...mmmh, I should't be happy me too !!!

you have to refer to the star schema, in which you have a fact table (in other words, your KFs) and dimensional tables (where you have to assign your char)...

but it's not so easy and you should have some datawarehouse fundamental...

take a look here:

http://help.sap.com/saphelp_nw04/helpdata/en/4c/89dc37c7f2d67ae10000009b38f889/content.htm

give me your e-mail !

Bye,

Roberto

[Former Member]

Ralph Kimball, Bill Inmon have articles on dimensional modeling, and if you search SDN and the web you can find more. (Seems like we need a wiki capability on this forum so we can build this kind of collective set of links). The topic is very broad and SAP's implementation of Star schema goes beyond the norm. I always like to try to diagram SAP's extended star schema when I get question from someone wanting to know why they can't just use Access and an ODBC connection to the BW and write their own queries.

It can be helpful to understand some of the differences among the different DBs that BW can be built on. You will often read about features in BW, and only later find out that your DB does not support them. Things like partitioning, bitmap indices, etc., are not the same or don't exist in some of the BW DB platforms and this might influence design.

I have not seen a good collection any where of the differences, and perhaps if I get some time this summer, I'll have to document what I have learned (or at least think I have learned).

[former_member184494]

Raja,

I would also like to add here that you should mention the extended start schema and should be aware of its benefits and disadvantages. ALso some reading on Dimansional Modelling and Datawarehousing is required. Coming from a DW context I wuold advise you to read Dimensional Modelling from the IBM Red Book available for free on the net , it is considered by many as a bible for Datawarehousing. Go thorugh the same and then through BW Modelling and you will get the bigger picture.

Arun

[Former Member]

if you guys want to download it...

http://www.google.it/url?sa=U&start=6&q=http://www.doc.ic.ac.uk/frank/kmt/redbook.pdf&e=10313

Bye,

Roberto

[Ashwin]

Hi raja sekar!

Welcome to BW forum....

your question has a very broad scope but i will try to put what i will do in modelling a cube with minimum steps.

it all starts with identifying the key performance indicators of a company.. then you identify the busniess objects related to this for example take the KPI sales. for this KPI the relavent business objects that might be intrested for management are customer , material , region... and so on. now you have to decide what are the describes these objects for example we may find some attributes which can give more detailed information. after identifiying these all we have to create a starschema for this requirements by using bubble charts and various other tools. here we need to identify the reporting requirement i.e. how to present this information and who requires this.

once the star schema is with us we can design the cubes dimentions and factable contents. aftet this we have to identify where we can get this data and what are the converstions required to load this data into the cube . at this point we need to consider the reporting requirements as we are going to present this contents in reports. then we create info objects, and cubes and other elements int he dataflow process. We have to determine the dataload process and its delta mechanism. here we may require to create process chains for regular uploads.

after loading data we create reports based on the identified requirements. during this process we may also require to deside on authorization aspects of the reporting.

these are all the aspects of modelling a cube and reporting from it. I hope other experts present in this forum will make nessesary corrections if i am wrong any where .

hope this helps you in having and overview of modelling cube.

with regards

ashwin