Continuing in the SAP World Tour in Toronto today, I went to a breakout innovation session on NetWeaver Business Warehouse (BW) and HANA, with Steve Holder from their BusinessObjects center of excellence. HANA, in case you’ve been hiding from all SAP press releases in the past two years, is an analytic appliance (High-performance ANalytic Applicance, in fact) that includes hardware and in-memory software for real-time analysis of non-aggregated information (i.e., not complex event processing). Previously, you would have had to move your BW data (which had probably already been ETL’d from your ERP to BW) over to HANA in order to take advantage of that processing power; now, you can actually make HANA be the persistence layer for BW instead of a relational database such as Oracle or DB2, so that the database behind BW becomes HANA. All the features of BW (such as cubes and analytic metadata) can be used just as they always could be, and any customizations such as custom extractors already done on BW by customers are supported, but moving to an in-memory provides a big uplift in speed.
Previously, BW provided data modeling, an analytical/planning engine, and data management, with the data storage in a relationship database. Now, BW only provides the data modeling, and everything else is pushed into HANA for in-memory performance. What sort of performance increases? Early customer pilots are seeing 10x faster data loading, 30x faster reporting (3x faster than BW Accelerator, another SAP in-memory analytics option), and a 20% reduction in administration and maintenance (no more RDBMS admins and servers). This is before the analytics have been optimized for in-memory: this is just a straight-up conversion of their existing data into HANA’s in-memory columnar storage. Once you turn on in-memory InfoCubes, you can eliminate physical cubes in favor of virtual cubes; there are a lot of other optimizations that can be done by eventually refactoring to take advantage of HANA’s capabilities, allowing for things such as interfacing to predictive analytics, and providing linear scaling of data, users and analysis.
This is not going to deprecate BW Accelerator, but provides options for moving forward that include a transition migration path from BWA to BW on HANA. BWA, however, provides performance increases for only a subset of BW data, so you can be sure that SAP will be encouraging people to move from BWA to BW on HANA.
A key message is that customers’ BW investments are completely preserved (although not time spent on BWA), since this is really just a back-end database conversion. Eventually, the entire Business Suite ERP system will run on top of HANA, so that there will be no ETL delay in moving operational data over to HANA for analysis; presumably, this will have the same sort of transparency to the front-end applications as does BW on HANA.