Re: SAP BW compared to Essbase/SQLServer/Oracle as a Enterprise Data Warehouse & OLAP application

Hello Amanda,

having lately had a look on all major multidimensional and relational OLAP systems I can just agree to all the statements regarding BW:

> > -difficult to use
> > -went through long implementation cycles
> > -performance was poor and scalability issues
> > -maintaining the application was expensive since SAP BW and ABAP
> > consultants are needed for ongoing maintenance.

you stated. At the moment there are only three reasons to implement SAP BW: 1. Your source data originates from SAP R/3 (at least 85 %). 2. Your users only know SAP R/3 reporting and are not "spoiled" by front ends from Cognos, Hyperion, Oracle, etc. or there are mainly (3) users receiving reports delivered by SAP's Business Explorer Web Application Builder (which is quite good).

Why 1? Because SAP BW offers predefined extractors to extract data from R/3. Be careful though since any additionaly created field in R/3 won't be covered by the extractors and need to be created manually. If you won't employ BW there will be no way around an ETL tool and consultants who know exactly where and how they get the data from (using routines that use the R/3 application by creating ABAP code through the ETL tool). The question is where you want to put the effort: In data extraction from R/3 or in maintaining your warehouse application server, the latter called BW. The options you stated with the ETL tools and their analytical apps form one scenario. Using the ETL tool, filling a database and employing separate front end tools is another. You are also comparing multidimensional and relational databases: In my oppinion you should see if you need the first or the latter to meet your requirements. This depends largely on the data volumes you want to handle. Data volume, #users and your required performance influence the decision which system and what platform to use. "Neutral" warehouse vendors like SAS, IBM, Microsoft or Oracle could do. Some of them even have "ETL" functionality built in which is quite good, like SAS or Oracle.

Regarding the front ends: There are indeed 3rd party front ends available for BW. The problem is just that the ODBO implementation is very often different from SAP's (although they have certified interfaces). OLAP BAPI gives less hazzles but only two vendors are certified. It is not correct though, that programming reports in BW needs any ABAP at all. There is a query builder you can use to create the report you want (in Excel or for the Web). The use of Business Explorer Analyzer is just not as comfortable than using Hyperions Excel Add-in or other vendors'.

> > RDBMS and the way SAP BW implementes aggregates is not like Oracle's
> > materialized views but copies the same data in the base cubes into a
> > number of additional star schemas so if a cube as 50 aggreagates these
> > 50 aggregates are like 50 small star schema cubes with the data in the
> > base cube copied to these aggregate tables as well with each build of
> > the BW cubes.

This is not a hundred per cent correct. SAP can use materialized views. And aggregates always look like small stars in relational data bases.

Apart from this these aggregates have indexes as well. We
> > looked at some of the SQL generated by SAP BW reporting solution which
> > is nothing but a Visual Basic for Application (VBA) addin to Microsoft
> > Excel the SQL queries for even simple tasks where long SQL selects
> > with so many joins.

You are right: SAP BW needs significant more joins to query an InfoCube than other tools. This results from its complex structure (texts, hierarchies, attributes, etc.) and slows down performance even if they use Oracle's star join.

 Apart, from this we were told
> > that despite the claim of SAP BW of hub and spoke data warehouse if we
> > need to take data in SAP BW to another data mart or data warehouse we
> > need to pay license to SAP for extracting our own data in SAP BW ??

This is true, ridiculous and shows how nervous the SAP people must be that someone uses a "fast" and "easy-to-use" multidimensional database for DSS or even just use BW as an extraction tool to get data out of R/3.

> >
> > Precisely I am wondering why not Oracle 9i OLAP or SQL Server 2000 or
> > Oracle/SQL Server with Hyperion Essbase/Cognos. Some of the SAP BW
> > consultants claim SAP BW is better than Business Objects, Cognos or
> > Hyperion.

There is exactly one situation where SAP's frontend Business Explorer is "better" than most other tools: When using BW as a data source and employing Business Explorer Web Application Builder for a Web based information system.

If you look at relational databases I would recommend to have a look onto 9i. I am not related to Oracle but their concept of integrating multidimensional OLAP into the RDBMS is far beyond IBM's who are just beginning to implement some OLAP functionality into DB2. Regarding your data volume NCR teradata or Microsoft SQL Server could be an option, too. And not to forget there are Sybase, Sand Technology and others too, not talking about the multidimensional data bases like Applix iTM1, MS Analysis Services, Essbase, ...

You were asking if SAP BW is a EDW. This is hard to answer and depend mainly on the amount of data sources you want to use for building it. SAP BW is a more convenient way to report data stored in R/3. It provides all the attributes and things you know from R/3 from Variables to texts, to (time dependent) hierarchies, to user settings, etc. In any case it will need significant more resources to implement and run a BW than to implement your solution on a neutral platform. In the latter case the only problem you have is to get the data out of R/3.

Kind regards, Received on Thu Jan 16 2003 - 11:25:17 CST