BACK

How fit is your SAP BW?

We have analyzed and benchmarked 150+ SAP systems in the world. Our overview shows that roughly half of the systems have between one and five terabytes. Given the current big data discussion, we ask ourselves where does the Big Data topic arise in context of SAP BW, given that most of these systems may be considered not that big.

No matter the system size there has arisen a number of areas that you should have in mind when managing your BW smartly.

 

DATA VOLUME

In our opinion an important indication of smart SAP BW management  is not just the size of the system, but the data volume that is to be requested and used in queries, as less than a third of the data is found in the reporting layer in an average system.

 

Furthermore, more than 40 percent of the system usually consists of temporary and miscellaneous data. Miscellaneous data means everything apart from core, cube, ODS, PSA data and change logs. We often find that SAP BW statistics charts as some the biggest volumes of data in the miscellaneous category. And, this data offers a huge potential for housekeeping.

 

DATA GROWTH

The current size of a BW system does not necessarily allow you to conclude what challenges may be facing us in the future. But, an important indicator is data growth; an average SAP BW system grows by an amazing 30 percent per annum and  makes growing pains inevitable. Potentially a 5 Terrabyte system two or three years later, will have doubled in size and become a 10 Terrabyte system.

 

On one hand, the costs for data storage technologies are decreasing and nn the other hand, the current developments with demands for faster data processing found in the likes of in-memory computing are pushing the price of a gigabyte back up again.

 

QUERY PERFORMANCE

Growing data stocks may not only lead to an increase in TCO (total cost of ownership), but there are  challenges focusing towards BW performance, as well. On average a BW system has a weighted query step runtime of 14 seconds, and have 57 read data sets per shown line. However what we have noticed in the top performing systems have a query that is comparable in size have a runtime of approximately three to five seconds.

 

To gain such a performance optimization in query runtimes like that in  the top performing SAP BW systems is actually possible in the average BW system while using existing hardware; they just need an improved understanding as to how to approach their BW system and it’s data.

 

How do you handle this in your system?

From what we have seen, SAP BW systems differ greatly in size and performance. In particular, performance can quickly lead to a great deal of frustration for business users. Switching technologies does not guarantee a notable increase in BW performance, as much as it doesn’t make a bad data model better. There are alternative ways to bring the SAP BW system up to scratch, however knowing where to start and understanding what is the most suitable approach for a good BW strategy is a proven path to follow.

 

Stiebel Eltron, one of the leading manufacturers of domestic and system technology, optimized their SAP Business Warehouse through the use of the DataVard BW Fitness Test. Both the detailed analysis and the benchmarking resulted in clear actionable recommendations on how to optimize system performance and decrease data volume:

 

  • Query runtimes reduced from 5min to 12sec
  • Data Volume reduction by 1/3
  • Buidling internal know how: after 1,5 year still improving the query runtimes

 

BW Fitness Shape Up at Stiebel Eltron

9 comments on “How fit is your SAP BW?

  1. Honing Data Science on

    SAP BW can be used for data science projects as well.
    1)in-house SAP BW developers and business analysts can build invaluable data science skills at minimal expense and disruption
    2)existing BW BEx queries are the natural starting point to deliver value through predictive analytics
    3)A rich library of over 8000 open source R-packages, used by millions of developers and data scientists, can be used to address practically any business problem
    4)industry standard predictive modelling language, R, can run 10-100x faster on live SAP BW data using TIBCO Spotfire

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *