Moving from BW-IP to SAP HANA optimized planning with the new Planning Applications Kit

In-memory technology is a success story within SAP NetWeaver BW for many years. The introduction of SAP NetWeaver BW accelerator provided a new level of reporting performance and became role model for others. Since then there was a strong desire to significantly accelerate planning use-cases with this technology, too. The value proposition can be summarized to:

• Improved plan quality (allow more simulations cycles)
• Improved user experience (provide better response time)
• Improved plan accuracy (process higher data volume) 

In general it is the mass data operations that benefit the most from in-memory technology: in reporting it naturally is the aggregation, in planning the disaggregation. But within planning there are much more mass data operations and every planning function is a candidate.

However for a significant performance benefit, mass data operations need to stay within the data layer completely, including data read, calculations and write-back.  BWA did not provide the durability of an acid-compliant database and as such was a secondary store which could not manage written-back data. With SAP HANA the same technology becomes available now with full acid-compliance.

To understand what it means to have the complete operations in HANA, let us look at the processing on a classical database (the width of the arrows describe the volume of data transferred):

First the data is read into a local cache in the application tier. There it is exposed into the plan session which is used to feed both, the BEx query for the end-user and the calculations in planning functions or disaggregation in the query. The calculations are tightly bound to different other components: the metadata of the plan application, constraints like characteristic combinations that the calculations must not violate and the delta buffer that contains the pending changes. These buffered deltas together with the locally cached data feed again the plan session. Finally the deltas in the buffer are written back to the database upon a save-command. All this is handled in the application tier for classical databases.

With HANA optimized planning all steps, data read, calculations and write-back, are done in HANA completely. The components remain the same: 

The plan session orchestrates the data flow between the physical data indexes and the consuming BEX query, the calculations in planning functions or disaggregation in the query. The data is read via projections into the level of aggregation demanded. The calculations are applied and the result written back into a delta buffer within HANA which is subject of further data requests. With this all mass data operations remain within HANA and only query relevant data and meta-data is exchanged between the application tier and HANA leading to significant reduction of IO-costs. In addition the columnar storage and parallel processing provide superior performance.

As a great benefit of this design the complete user experience remains untouched. This is true for the end-user clients (e.g. BEx suite, Advanced analysis for office) as well as for the modeling UI (ABAP planning modeler) and all existing BW-IP models. I.e. there is no need to migrate BW-IP scenarios to run on HANA. Adjustments might be considered though to optimize the HANA usage since not the complete BW-IP feature set can be executed in HANA today (see note 1637199). The other way around, all capabilities offered in HANA are available in the ABAP runtime as well. This allows to toggle between two operation modes of BW integrated planning on HANA:

 Coming from an existing BW 7.x installation (A), the upgrade comprises a simple upgrade to BW 7.30 SP5 on the existing database and a subsequent database migration to HANA. Here BW-IP leverages the SQL-interface of HANA leading to superior read performance. Plan calculations are executed in ABAP (B), still. Their execution in HANA can be enabled by activation of the planning applications kit, activated by flipping a switch (see note 1637199) (C). The planning applications kit leverages the calculation and planning engines build into HANA to process the plan calculations in the best possible performance.  This way the planning applications kit combines the feature-rich capabilities of BW-IP with the superior performance of SAP HANA.

Finally let me summarize the relation between BW-IP and the planning applications kit (PAK).