The Data Product Generator and query templates

The Data Product Generator is the new piece most BW teams have not used. It lets BW keep running while its data becomes available to SAP Datasphere, SAP Analytics Cloud and SAP Databricks as data products. This lesson explains how it works, what it supports, and how to test it.

How it works

1. Meet the prerequisites: BW in the private cloud edition, on SAP HANA, at a supported release. SAP Learning lists BW/4HANA 2023 SP0 or higher, BW/4HANA 2021 SP4 or higher, and BW 7.50 SP24 or higher.
2. Create a subscription on an InfoProvider: in BW 7.5 with transaction RSDPS in SAP GUI; in BW/4HANA with the Data Subscription app in the BW/4HANA cockpit or in the BW modeling tools. You can project columns and filter rows.
3. Activate it: a local table (file) is created in a BW inbound space in Datasphere, backed by SAP HANA data lake files (an inbound buffer, a merge task and a delta table).
4. Run it: BW pushes the data into the inbound buffer and the merge task consolidates it into the delta table. Delta runs skip BW requests already replicated, where the InfoProvider supports delta; in BW/4HANA, subscriptions can be scheduled in process chains.
5. Publish it: in the Datasphere Data Sharing Cockpit, create a data provider profile, define the data product, and list it so it gets a Delta Sharing endpoint and appears in the catalog for consumers such as SAP Databricks.

The Query Template Generator

SAP's materials also describe a Query Template Generator that transfers BW query metadata into Datasphere and translates it into an analytic model with a fact source and reusable dimensions, built on BW data products. SAP's reference architecture marked it as planned, and SAP Learning's BW transformation course has a unit on it; check its availability for your tenant. Even when available, treat the generated model as a starting point: restricted and calculated key figures, conditions, exceptions and hierarchies need to be checked one by one, because a template that looks complete can still differ in a single formula.

Operationally, subscriptions become part of the BW schedule. In BW/4HANA they can run in process chains after the loads they depend on; in BW 7.5 they are scheduled from RSDPS. Noor adds three checks to the daily operations routine: did each subscription run after its source load, did it finish without errors, and do the row counts of the delta run match the BW requests loaded that day. When a subscription fails, consumers of the data product see yesterday’s data without any error of their own, so the monitoring must sit on the BW side and alert the owners of every consumer.

Not every InfoProvider should become a data product. Exposing data is cheap to start and expensive to support: every product needs an owner, reconciliation, access design and monitoring. Contoso starts with three: the billing aDSO (for the warranty prediction work in SAP Databricks), the margin CompositeProvider (for the new SAC margin story while the Datasphere rebuild is under way), and the inventory snapshot InfoCube (to start building history in Datasphere early). Each has a named consumer and a planned end: when the rebuilt model replaces it, the BW data product is retired too.