The BI Accelerator is a new approach for enhancing SAP Net Weaver BI performance based on SAP's search and classification engine TREX, and on preconfigured hardware delivered by SAP hardware partners. It is packaged as an appliance for use with SAP Net Weaver Business Intelligence (BI) and provides enhanced performance for OLAP in an Enterprise Data Warehousing IT scenario.
A TREX aggregation engine for processing structured business data enhances the performance. The data from the BI Info Cubes is indexed in the BI Accelerator and stored as TREX indexes in its storage subsystem. The BIA indexes are loaded into memory and used to answer OLAP queries entirely in memory. The BI Accelerator clearly reduces the response time, especially for large data volumes. SAP Net Weaver BI customers adopting the BI Accelerator can expect significant improvements in query performance through in-memory data compression and horizontal and vertical data partitioning, with near zero administrative overhead. Since BI Accelerator is delivered to the customer as a preinstalled and preconfigured system on dedicated hardware as a BI Accelerator box, the installation and initial configuration has been done by the hardware partner and no additional administrative tasks need to be done by the customer for the first usage of the BI Accelerator.
Normally, optimization of query response times requires some form of prior aggregation. The BWA, however, performs aggregation at query runtime. With the BWA, all query processing is handled in memory. There are no disk accesses.
The BWA first indexes each InfoCube to be accelerated. This index structure is highly compressed and loaded into memory whenever the query requests the data. Indexing is performed once and the indexes are updated automatically as required.
When an index is in memory, user navigational steps from the query are managed from memory. There is no need for disk access as users navigate through queries.
The BWA processes any incoming queries against indexes in memory, using fast join and aggregation algorithms based on efficient integer coding techniques. The results are then passed to the SAP BW analytic engine (OLAP) and returned to the user.
To create the indexes, database tables are split vertically and stored by column, not row, and the indexes can be split horizontally for highly parallel processing on multiple processors.
The BWA uses smart compression, which is based on dictionaries and integer processing, and is optimized to run on Intel Blade Servers. The Blade Server infrastructure allows for a dynamic assignment of hardware resources to ensure high availability and good load balancing.
BW System and BWA:
The following figure depicts the BI Accelerator architecture and its releationship with the BW-system
The BW Accelerator is installed on a preconfigured blade system. A blade system consists of hosts in the form of server blades. The server blades are connected to central disk storage. This is referred to here as a file server, regardless of the underlying hardware. The special feature of a BW Accelerator installation on a blade system is that the BW Accelerator software can be stored centrally as well as the BW Accelerator data. This means that the software will be installed only once on the file server. Maintaining the system is efficient because you only have to implement software updates once.
All server blades on which BW Accelerator is running access the same program files. However, each server blade has its own configuration files. The configuration files in the directory <TREX_DIR> are only used as templates. A script creates a seperate subdirectory for each server blade and copies the configuration files to this subdirectory.
Benefits of the BWA:
- Faster query processing and response time. The BWA is designed to filter and aggregate very large data volumes in a very short period of time. This is a significant advantage compared to the classic relational database system.
- Lower maintenance costs:
- BI Accelerator eliminates the need to create relational aggregates.
- BI Accelerator results in less planning and tuning on the part of DBAs.
- High potential scalability. As demands grow, system scales up by adding blades.
Shortcomings of the BWA:
- Currently the data source for the BW accelerator can only be an SAP InfoCube. It does not work with other infoproviders such as a DSO.
- In a productive environment, a one-to-one relationship is mandatorly required between the BW-system and the BWA. The use of more than one BW-system with one BWA is not part of the original BWA concept (performance, distribution of data, possibility of restarting the BWA), since it does not allow the synchronization of data across several BW-systems.
- The data retrieved from the BWA is transferred using a Remote Function Call (RFC). This interface is not designed for transferring mass data. Hence the data-amount retrieved from the BWA must not be too big.