How Planning Application Kit (PAK) can be useful in improving the performance of your planning application.
I had been working on a planning application built using BW Integrated Planning for one of the clients. That application had extensive usage of the “Distribution by Reference Data” standard planning function. It also used the FOX functions for distribution customization. Apart from that, it used other standard planning functions like “Delete” and “Copy”. The application was built on BW7.5 on HANA.
The issue we were facing was that the planning sequences executing these functions were taking a long time and the performance was exponentially worse with increased volume.
In order to resolve the problem, we enabled the Planning Application Kit (PAK) for the HANA In-Memory execution of planning functions.
We faced some issues after enabling the PAK as follows;
- There was a data slice of type exit on top of the planning provider due to which it wasn’t allowing to run it on HANA In-Memory. Hence, we had to deactivate the data slice.
- Some of the planning providers were not HANA optimized. Therefore, we converted them to HANA Optimized.
Once it is ensured that the system is now allowing the function to execute on HANA, we performed the following steps to analyze the performance improvement.
For each month and region:
- Data was selectively * deleted
- Jobs were triggered with PAK switch OFF
- After successful execution, data was selectively * deleted
- Jobs were triggered with PAK switch ON
- Runtime was noted for both the scenarios (PAK switch ON & OFF)
* Selective deletions were performed to maintain an equal number of record processing in both scenarios.
We were able to achieve around 90-95% improvement in the execution of standard planning functions viz. “Distribution by Reference Data”, “Copy”, and “Delete”. However, there was some loss of 1-2% in the execution of FOX functions because they failed to execute on HANA In-Memory due to custom code.
In order to overcome the issue of FOX functions, we converted them into custom planning functions using AMDP (ABAP Managed Database Procedures) classes.
We were able to achieve very good results (Around 90-95% improvement) even on FOX function converted to AMDP.
Hence, we were able to execute the jobs 90% faster using PAK. Another observation was that the performance improvement was exponentially proportioned to data volume. Regions with less volume of data didn’t see the improvement as compared to the regions with huge data volume.
In Nutshell, Planning Application Kit (PAK) is a great way to utilize HANA In-Memory execution capability for the planning applications built-in BW Integrated Planning.