Professional Documents
Culture Documents
Februari 2008
The struggle for Management Information from SAP R/3.
Many organizations are struggling with their Management Information. There are many needs and
many solutions to provide this information and it is exactly this that makes the problem so complex. In
large companies there are many requirements that need to be met and, quite often, huge data volumes
are involved as well.
Looking at all the solutions that are offered, it is not a straightforward choice pick any of the solutions on
offer. Typically all offerings have their strengths and weaknesses. Even knowing and understanding
those can be a lifetime mission. Although the market for MIS solutions is consolidating because of
mergers and acquisitions there are still many options
One of the most apparent omissions in R/3 is the lack of reporting capabilities for management
purposes. The way the system was built is exactly what makes R/3 a difficult system from a MIS point
of view. There are thousands of tables used by R/3 that may contain relevant information. Also the
limitation that you cannot access the underlying database directly makes things complicated.
So at first sight SAP BW seems like a no-brainer. Being from the same manufacturer as R/3 it must
'understand' the R/3 system better than anything else. Still, there are many companies that are
struggling to implement SAP BW. Why? There are several reasons for this and it's the combination of
factors that hampers a smooth implementation. We will try to explain this below.
Well, yes and no. The way R/3 processes data is much different from the way a Data warehouse is
processing data. Although SAP made good efforts in adapting the kernel to be able to handle these
bulk operations, it is still using a procedural system to deal with bulk operations on (potential) huge data
sets.
Are there no synergy advantages because both systems use the same programming language and
infrastructure setup?
Again this looks like an advantage but it is actually not. There are fundamental differences between the
setup of an OLTP (online transaction processing) system and a reporting system. This means that the
programming language ABAP is rooted in the procedural handling of data used in a OLTP. It involves
the selection of data into a temporary structures and then perform the required operations. Typically
systems used in MIS handle data in the format of streams where you modify data by operating on the
data stream. The differences become clear when you are using huge volumes of data, the data stream
architecture typically has a smaller memory footprint, making it faster and much more scalable.
The proof of this is the release of the BI Accelerator. This is a hardware based (technical) solution to
improve the performance of SAP BW. The BI Accelerator deserves an article of it’s own so for here it is
enough to state that it is an expensive stopgap bringing no improvement to architecture of BW.
Unfortunately most organizations are not paying enough attention to this phase of building a data
warehouse. By not carefully designing your model you are building legacy from scratch as it will be a lot
of work afterwards to change models to accommodate for new needs. SAP has this very tempting offer
inside SAP BW called 'Business Content'. It basically gives you a head start in setting up your data
warehouse. By simply activating the content you get ready defined structures to build your reports on.
So (again) if that is the case, why do most BW project take so much time and effort?
The first option means a lot of work, all definitions have to be put in place, all structures need to be build
as well. The second option looks much more promising. Early flavor results can be made available in a
relatively short period of time.
Trying to change the pre-configured definitions makes it apparent why SAP BW is a very time
consuming system to build. Changing structures in BW is not a trivial activity. There are many steps
involved and quite often data needs to be reloaded into the system. As the system tries to prevent data
corruption by guiding the designer through a carefully designed procedure for changing objects it also
limits the number of 'tricks' an experienced DBA can use to migrate data structures.
By relying on an system to maintain the setup of your data warehouse, you are no longer able to
change things 'manually'. Although this seems like an advantage, daily practice proves it has a serious
and negative impact on progress. Most changes will take a long time because of the inefficient way of
working. You may consider this being the trade off for having a system that protects you from messing
up your data warehouse. However when your R/3 system is heavily customized and you are dealing
with large data volumes this will become a problem.
Looking directly in the BW database can be accommodated but will only lead to confusion as the
metadata, required to understand the business data is not available in the database unless you use the
BW front end. This is actually a shame as (for example) Oracle database provides many ways to
include metadata into the database itself (meaningful field/table names, field comments, etc.).
Maintaining a data warehouse will always require manual interventions at some point whether you like
it or not. The 'distance' BW sets between the storage structures and the administrators will lead to much
frustration and delays. It is similar to the cars of today. It looks like everything can be diagnosed and
solved via the computer, but in the end you still need to get a down and dirty to the fix a problem.
Although the above way of working also brings you SAP BW's equivalent of the 'undo' button it does put
a significant toll on performance when data volumes are high. It also adds additional tasks to the
housekeeping department. The number of checks you need to perform in a fully grown BW system are
much bigger when compared to a database operated data warehouse.
As described earlier BW comes with a pre-packaged definition set. This also leads to significant storage
needs for data that is not used. Because SAP uses generic extractors to retrieve data from R/3 you will
end up with much more then what you need.
You need to be very clear on the costs and benefits of a BW system before you start! A good start is to
look back on how you experienced the implementation of R/3 itself. Most likely you reached a point
there when you thought 'how much more is this gonna take'. Rest assured for BW it will be no different.
It is this experience that has had large impact on Arno’s way of working. He has been very successfull
implementing solutions for SAP. Never underestimating the effort always open and upfront to his
clients, making sure that they know what they have to know.
Arno was amongst the first that joined NewFrontiers and as such has had significant impact on what the
company is like to day. Key thing put your money where your mouth is! He has built up significant
experience building data warehouses on SAP ERP for organizations like Shell, Epson and Unilever.
Most of his clients wanted Arno to stay for many years which is where and why he has build up so
much in depth knowledge of the different techniques and tools in place like SAP BW (today called
Netweaver BI).