Via Content in SCN
While 2012 was all about closing BW on HANA deals, 2013 has been all about BW on HANA implementations. This article describes the journey of my recent BW on HANA success story at a retail customer in Nordics.
Customer Use case:
The customer has implemented SAP ERP and SAP BW systems. Recently, they have incorporated point-of-sale data analytics in the BW system. In brief, the use case was as follows:
8PM: Retail stores across the country closes.
9PM: Point-of-Sale (POS) data is ready to be loaded in the BW systems.
9PM-3AM: Data should be loaded in the BW system and KPI calculations need to be performed in the BW system.
3AM: BW system must be ready by this time to load the data back in the ERP system. This aggregated data with the KPI calculations is then used in the ERP system to execute more application logic.
7AM: All the processing must be complete by this time in the ERP system so that store managers have relevant information available before sto res open.
Challenges
The challenge with the above timeline is that there is only a six hour window to:
Customer Use case:
The customer has implemented SAP ERP and SAP BW systems. Recently, they have incorporated point-of-sale data analytics in the BW system. In brief, the use case was as follows:
8PM: Retail stores across the country closes.
9PM: Point-of-Sale (POS) data is ready to be loaded in the BW systems.
9PM-3AM: Data should be loaded in the BW system and KPI calculations need to be performed in the BW system.
3AM: BW system must be ready by this time to load the data back in the ERP system. This aggregated data with the KPI calculations is then used in the ERP system to execute more application logic.
7AM: All the processing must be complete by this time in the ERP system so that store managers have relevant information available before sto res open.
The challenge with the above timeline is that there is only a six hour window to:
- Get the data transferred from the SAP ERP system to the BW system.
- The process chains in the BW system must be run where the data is loaded from the PSA to DSO and Infocubes and propagated through various complex application logic layers.
- Various KPI calculations are performed in the BW layer issuing many look-ups
We installed BW on HANA and connected it to the source ERP system. There was an existing BW on a traditional DB (BW on ODB) system already existing. This arrangement made it possible to load the data in parallel in the two BW systems and compare the performance numbers.
We loaded same number of records on both the systems and ran the process chains and recorded the below numbers:
During the realization, our BW on HANA expert found possibilities to optimize customer's current ABAP code and BW modeling. Just with these optimizations, customer may roll out their existing implementation to a nother smaller market without any issues in near future, though they would need HANA as soon as they want to include bigger markets.
As proven during the performance tests, there were huge gains in DSO activation times. Usually, in a multi-layer architecture, DSOs feed other DSOs, which was also the case here. The activation gains for each DSO got multiplied and this resulted in great performance improvements.
We also demonstrated loading more than 600 million records in the BW on HANA system without any issues with only 3 parallel jobs. By this we showed that, the system can handle heavy loads without any issues.
Next step at the customer would be to bring in data in the the BW on HANA system from more countries and be able to run the process chains in the BW system with hundreds of millions of rows empowering the store manager more and more.
No comments:
Post a Comment