Tcodes: RSDDV, RSADMIN
BI Accelerator is a separate server which runs on TREX (Text Retrieval Extraction). The BI-Accelerator is a an addition to SAP BW 7 version and is now becoming obsolete with the advent of HANA.
BIA serves the purpose of putting some of the data in the memory whereas HANA puts all the data in memory.
Many companies have opted for BIA when it was launched and it has given impressive performance with typical improvement factors between 10 and 100.
Long running Sales/AR/Finance queries have improved drastically when the cubes on which they are based are indexed on BIA.
Based upon the TREX search engine for unstructured data, it applies its technology to the structured data stored in the infoCubes of an SAP BI system.
The BI-Accelerator achieves these performance gains by applying a vertical decomposition algorithm to data and storing and indexing data by column (fields) instead of rows or tuples as in traditional database.
An efficient compression algorithm reduces the size of the indexes by a factor up to 20, allowing BIA to keep indexes in the main memory rather than having to retrieve indexes from disk space all the time.
BIA runs on a separate, highly parallel blade server. It has parallelized indexing and retrieval algorithm.
BIA is helpful where aggregate fails wherein there is large data volumes (more than a billion records) and adhoc query requirements.
Procedure to maintain BIA for a cube-
Select your cube, right click and select ‘Maintain BI-Accelerator Index’
or go to the tcode ‘RSDDV‘
BIA index is basically a structure containing replicated data and indexes (1 index per table).
Indexing: Transfer of Infocube data to the BIA server and processing and compressing it into the BIA index.
- In BI 7.0, InfoCubes can have both aggregates and BIA
- From the data maintenance perspective, the BIA is very similar to aggregate rollup
- Data is sent from the file storage to main memory
- Processing of data is done in the main memory
- Data will be aggregated on the fly
- BEx query fetches data from the main memory
Note: Main memory consists of blades ( two to five blades) and each blade consists of six to seven processors (CPU) and each CPU consists of 20-30 GB RAM
All the infoCube data is placed in the RAM (Main memory)
Change run (Attribute change run, hierarchy change run) and rollup activities will be maintained by BIA.
When you execute your query, it fetches data from main memory storage and delta storage. After 3 to 4 days, BIA merges the delta storage to the main memory using a background job.
If there is a structure change for the infoCube, you need to rebuild the indexes for that infoCube. InfoCube containing more than a billion records need to be on BIA, no need to maintain aggregates for them.
You can have only one BIA server for a BI system. Master data is stored in the BIA. All the indexes are maintained by system ID. Each infoCube can have only one BIA index.
Note: BIA will not support non cumulative key figures. Thus, an infocube containing non cumulative key figures e.g. 0IC_C03 cannot be on BIA.
BIA is available as an appliance. The hardware partners are HP, IBM, FUJITSU and SIEMENS. The CPUs for these BIA appliances are made by Intel.
RSADMIN is the tcode to maintain the RFC connection between the BIA and the BI server.