Author(s): Anil Kumar Puranam and Peesu Sudhakar
Company: Deloitte
Created on: 30th October, 2014
Authors Bio:
Sudhakar Reddy Peesu is an SAP Technology professional with more than 15 years of experience leading SAP BW/ABAP implementations, developing custom applications, and developing IT strategy. He is a PMP certified manager, has led BW/ABAP teams with hardware and landscape planning, design, installation and support, and development teams through all phases of the SDLC using the onsite/offshore delivery model. He has strong communication, project management, and organization skills. His industry project experience includes Banking, Consumer/Products, and Public Sector industries
Anil Kumar. Puranam is working as a Senior BW/BI Developer in Deloitte consulting. He has more than 10 years of SAP BW/BI/BO experience. He has worked on various support/implementation projects while working in consulting companies like Deloitte, IBM India, and TCS.
How to rebuild the BWA with BWA splits:
We generally build the BWA on the cube when we started the data loads. And there is highly possible that the data in the cube is minimal at the time of BWA initial fill compare to the deltas that we receive. If we do BWA init fill with less data, we would endup reorgnize or rebuild due to large data roll up in the delta laods.
To aviod this proble, while creating the BA , we can give the instruction to system to build the BWA witht the estimated data in the cube. So that system will allocate the enough BWA index space to the Info prociders by allocating space in different blades. The below screen shot shows how to rebuild BWA with splits.
Approach: When we index into BWA for first time i.e INITIAL FILL, we need to switch to “Own Setting” in “LOAD Admin” tab and go high value for the number of records . This will force the system to split BWA.
Estimate Runtime of Fact Table Indexing:
We can get this information from RSRV checks:
All Elementary Tests → BI Accelerator:
The system estimates the time required to fill the fact index. It uses the current parameter values for background and dialog parallel processing. The time taken is calculated from the processes available and the estimated maximal throughput of data records in the database, the application server, and the BIA server.
The calculated duration is an estimate; the load on the system, the distribution of data across block criteria and deviations during processing can all affect the actual time taken.
For ex:
Results after checking:
Estimate Memory Consumption of Fact Table Index
We can get this information from RSRV checks:
All Elementary Tests → BI Accelerator:
The system estimates the size of the fact table index of a BI accelerator index. In doing so, the system analyzes the data in the fact table and provides a projection.
For ex:
Results after checking:
BWA alert when the index size reaches 1billion.
- Program YBWA_ALERT_SEND
- Currently this program does, 2.1 billion limit check as well as Load Admin setting check (Own Setting Vs System Setting) from BWA end
- When there are exceptions in the above check the program, email sent out the exceptions to the email distribution entered.
- We can add the program to Process Chain to run them on daily basis to execute these checks.
Porgram is attached.
BWA overloaded alert:
Business Warehouse Accelerator (BWA) runs properly if we limit for the amount of data we can load into the BWA compare to thresh hold limit. The thresh hold limit is defined as below We should only load < “50% of the available blade memory” into the BWA.
- Report RSDDTZA_BIA_OVERLOAD_EMAIL can be used to check the availability
- We can create process chain to schedule a broadcast mail, for which we need to use report RSDDTREX_ADMIN_MAINTAIN to specify the e-mail
addresses. - Entry to maintain the email ids are BIA_NOTIFY_EMAIL_ADDRESS1 ,BIA_NOTIFY_EMAIL_ADDRESS2,… etc
Refer note 1767424 for more information.
Sample output:
BWA query performance metrics process
- BEx Query can be created as below on 0TCT_C03
- This query gives trend of BWA Read Time in Seconds per Million records.
- You can provide a date range and Multi providers as selection.
- In Addition to Time to read 1 M records, there are other measures such as time to read hierarchy, Time to Prepare Read, No of records
Read etc are available for display in this query
We can also use one of the Analysis from ZBWA_TOOLKIT program. We can select Index Usage Analysis and we can go with different selections provided there.
