07.02.2013 Views

Best Practices for SAP BI using DB2 9 for z/OS - IBM Redbooks

Best Practices for SAP BI using DB2 9 for z/OS - IBM Redbooks

Best Practices for SAP BI using DB2 9 for z/OS - IBM Redbooks

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

and right-click to select the Simulation Update tool. Additionally, you can use<br />

SM30 or ST05 to identify the ABAP and SQL used in expensive update and<br />

transfer rules.<br />

Loading data targets (ODS/InfoCube) takes too long<br />

For this condition try to load master data be<strong>for</strong>e transactional data <strong>for</strong> ODS and<br />

InfoCube. By creating all necessary SIDs and populating master data tables first,<br />

there will be significant per<strong>for</strong>mance improvement because you avoid expensive<br />

SID creation during transactional data load.<br />

In addition, try to delete be<strong>for</strong>e load, if possible. Per<strong>for</strong>mance is gained by<br />

deleting existing data be<strong>for</strong>e a complete replacement load. Deleting data from<br />

PSA helps to reduce PSA access times. Deleting data from InfoCube helps to<br />

reduce deletion and compression times.<br />

You can also influence per<strong>for</strong>mance improvement by parallel processing. For<br />

BW 2.x, it is not possible to have data packets/requests loaded into an ODS<br />

object in parallel. In BW 3.x, you can load data from different requests in parallel<br />

to an ODS activation queue. Transaction RSCUSTA2 can be used to control<br />

data ODS data load. Here you can define a maximum number of parallel dialog<br />

work processes, a minimum number of records per package, a server group <strong>for</strong><br />

load balancing, and so on.<br />

For non-reporting ODS objects, if possible, turn off the BEx Reporting setting.<br />

Load will be faster as Master Data SID tables do not have to be read and linked<br />

to the ODS data. You can do this by selecting RSA1 → InfoProvider, then<br />

double-check the ODS object and expand settings to turn off the BEx Reporting<br />

box.<br />

For initial large data load, there are significant database accesses to the NRIV<br />

table to fulfill a number of range requests. You can reduce application server<br />

access to the database by buffering the SID number range <strong>for</strong> InfoCube. This<br />

can be done <strong>using</strong> transaction SNRO to increase the number range buffer. After<br />

load completion, you may want to reset the number range buffer to its original<br />

values to minimize unnecessary memory allocation. Refer to <strong>SAP</strong> note 130253<br />

<strong>for</strong> more details.<br />

From the RSPC graphical view of the process chain, you can drop down to the<br />

SM37 job log to see more detailed activities. You can see whether there is index<br />

manipulation, and which tables are being processed.<br />

Regarding index manipulation, you need to ask whether this is necessary <strong>for</strong><br />

your particular loading scenario. Generally, index manipulation is effective <strong>for</strong> the<br />

initial full load of large data because of a high volume of insertions. For the initial<br />

load of very large data, consider dropping secondary indexes to improve loading<br />

156 <strong>Best</strong> <strong>Practices</strong> <strong>for</strong> <strong>SAP</strong> <strong>BI</strong> <strong>using</strong> <strong>DB2</strong> 9 <strong>for</strong> z/<strong>OS</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!