Introduction

The Business Requirement

Business applications normally store operative data in the form of business documents and master data. In SAP BW, DataSources and their extractors make it possible to access this data. DataSources provide a flat analytical view of the data in business documents, such as sales orders or purchase orders. They contains the business logic that derives an analytical view of the data from the transactional view. There are various types of DataSource: DataSources for transaction data, master data attributes, master data texts and master data hierarchies. Until now, DataSources have been used to replicate mass data from the operative system to SAP NetWeaver's Data Warehouse, SAP BW. Here, the data is integrated from various sources, consolidated and made available for OLAP analyses. OLAP analyses in BW are not based directly on DataSources though. They are based on InfoProviders. InfoProviders provide a view of a dataset from various, semantically related DataSources (purchase order data, customer data or product data for example). Scenarios for Operational Analytics hardly enjoy the benefits offered by a Data Warehouse though. The data from just one operative system should be analyzed here. Scenarios of this kind are kept simple in terms of extraction, transformation and loading (ETL) and data administration. They do not need any complex conversions or services like (Near-Line ) archiving. The additional costs that arise in connection with a Data Warehouse, maintenance and administration for example, are often inappropriate in application cases like these. On the other hand, the OLAP requirements and the quantity of the data to be processed can be unwieldy, making it necessary - in complex system landscapes for example - to support data extraction and data replication to a Data Warehouse and to consolidate the data there.

The Solution - Operational Data Provisioning

With Operational Data Provisioning, SAP NetWeaver provides a metadata concept that allows analytic query access for OLAP analysis in an operative system and replication scenarios (including an ETL service with a delta mechanism). Operational Data Provisioning is implemented in a modeling environment used together with the search and provides a metadata view in which a DataSource can be given analytical properties in order to define an Operational Data Provider (ODP). An Operational Data Provider can be used to access the data for the replication in various consumers (BWA or SAP BusinessObjects Data Services for example) and for the purpose of operational analytics. DataSources are not suitable for Operational Analytics, as they are too basic: A transaction data DataSource does not recognize the associated master data attributes, and the DataSource for master data attributes does not recognize the associated texts. Operational Data Provisioning uses ODPs here to allow semantically related DataSources to act as InfoProviders, so that the data is available to the Analytic Engine in an Operational Analytics scenario without the need for replication to SAP BW.

The graphic below shows how data provisioning works, including data replication using ODPs:

On the left side of the graphic, data provision for ODPs is shown. The data provisioning is based on the extractors that implement the business logic. Many SAP applications, especially the SAP Business Suite, provide BW DataSources with application extractors. These denormalize the application's entity relationship model and provide consistent data for the analysis. To support Operational Data Provisioning, DataSources are given ODP properties and are linked with one another. As well as the BW DataSources, there are SAP applications that provide data sources. ODP Modeling is integrated in a joint development environment for data provisioning for search and analysis. This means that uniform data sources can be provided for analysis and searches.

The right side of the graphic shows how - in addition to direct access - Operational Data Provisioning supports consistent replication of mass data. Accelerated access to data is made possible by using SAP HANA or the SAP NetWeaver BWA "out of the box". These make it possible to index ODPs periodically. Depending on the application extractor, it is even possible to perform real-time indexing with a latency of under an hour. Indexing is supported for DataSources and for other data sources too. For DataSources, it is also possible to replicate mass data using an ETL interface. Delta methods are fully supported here by the services in the delta queue. The delta queue allows multiple subscribers to access the data, thus separating the extractor from the various targets. The delta queue keeps the data compressed and makes it possible to monitor the data and correct errors.