SAP and In-Memory: Take Three

BAE encompasses transactional and analytic data. No other product or vendor does this. In the keynote, founder and chairman Dr. Hasso Plattner started off viewing a monthly sales chart in SAP BusinessObjects Explorer (product review here). Sales projection looked lower than target. In contacting the production manager, the production manager could review open orders from within SAP's Business byDesign; the transaction system was working off the same in-memory database as SAP BusinessObjects Explorer. Once the confirmed date was updated in the transaction system, the change was immediately apparent in Explorer (demoed, by the way, on an iPad. Integrating the operational and analytic processes opens up new applications. When I visited the Labs at Sapphire, I also saw a prototype of a customer segmentation application. The segmentation was done so intuitively and quickly, that the whole segmentation process was more iterative than normally possible. A de-duplication of records, that typically takes hours, took seconds.

The new BAE does not require BW, unlike BW Accelerator (BWA). SAP's previous in-memory approach only benefited customers who had committed to BW. In this regard, there is greater flexibility than with BWA, because star schemas do not have to be built first. The new engine promises to benefit existing Business Objects customers.

The BAE is an appliance; most other in-memory approaches on the market are software-only. HP was highlighted as an initial partner co-developing the solution, but IBM was also mentioned in a subsequent interview. IBM is a hardware partner for the BWA appliance.

The modeling of the in-memory model will be incorporated with the current business meta data layer or universe; so it sounds like both business users and DBAs will have modeling capabilities that today are often separate. More complex calculations can be done in memory.

The database will be accessible via SQL, MDX or Excel, so it is not disruptive to the application layer or front-ends. I still have to see this still to believe it.

Planning applications will leverage the in-memory engine bringing more flexibility and allowing for more iterations than currently possible. In theory, this would also provides more integration between BI and planning than other vendor solutions.