Folks from SAP BW world are familiar with the concept of “Extended Star Schema (ESS)”, which is SAP’s own implementing the classic snow flake model. Although the ESS had its advantages, it also had its drawbacks in terms of performance due to unusually large number of tables that come into play.
Image on the right depicts a cube with one expanded object/ This was the architecture from the day SAP BW was launched.
SAP BW optimized the data model along with introduction of HANA as the database platform. This optimized model, although can be applied for non-HANA databases too was not really promulgated (for reasons known only to SAP). HANA based models, apart for taking advantage of inherent advantages of HANA, tweaked the ESS model to bypass the Dimension tables, thereby linking fact data directly with SID tables. So, the fact data was one step closer to actual master thereby improving query execution times. Other redundant items such as aggregates, collapse were also removed thereby making the model simpler.
The next major move, the architecture of ESS is going to be transformed with the new B4HANA approach. The new approach has eliminated the concept of InfoCube and now has relegated all the modelling to a simple structure via Data Store Objects. The replacement for InfoCube is an object called Advanced Data Store Object, that can do the role of a cube as well as DSO.
The major shakeup here is the use of primary key instead of SID tables for reporting purposes. This is a simpler model than previous one with wherein all dimension keys are natural keys in the model.
Legacy objects are not supported anymore and tools are available to migrate the entire data flow(s) into new format. This jump from old architecture to new one is quite tedious and basically has impact on entire BW system. Although advantages are clear, adopting a drastic jump into new architecture will be challenging from all angles.