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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

– When not condensing all requests:<br />

i. Drop the secondary indexes on the F- fact table.<br />

ii. Call function module RSDU_INFOCUBE_INDEXES_DROP <strong>for</strong> the F-<br />

fact table.<br />

iii. Condense the requests.<br />

iv. Call function module RSDU_INFOCUBE_INDEXES_REPAIR.<br />

This speeds up the deletion of the request from the F- fact table, because<br />

the deletion of entries from the secondary indexes is very expensive.<br />

– Activate F- fact table partitioning.<br />

See <strong>SAP</strong> notes 860830 and 894045.<br />

► Tune the size of the bufferpool <strong>for</strong> the E- fact table index pages. If the<br />

bufferpool size is too small, it is the bottleneck <strong>for</strong> the insert into the E- fact<br />

table.<br />

The default bufferpool <strong>for</strong> all index pages is BP3. The specific setting <strong>for</strong> fact<br />

table index pages via the RSADMIN parameter is<br />

<strong>DB2</strong>_FACT_BPOOL_INDEX (see <strong>SAP</strong> note 536074).<br />

► When condensing many large requests:<br />

a. Drop the secondary indexes on the E- fact table be<strong>for</strong>e starting the<br />

condensing.<br />

b. Recreate the secondary indexes after all requests are compressed. Call<br />

function module RSDU_INFOCUBE_INDEXES_DROP and specify the E-<br />

fact table name.<br />

c. Call function module RSDU_INFOCUBE_INDEXES_REPAIR to rebuild<br />

indexes.<br />

Do not drop the secondary indexes if the E- fact table is already very large<br />

and the compression adds less than about 10% of new rows into the E- fact<br />

table.<br />

► Restrict the size of the requests.<br />

– Requests should not be larger than 10 million rows.<br />

– Compression of huge requests results in long-running uncommitted UR<br />

(unit of recovery) and long elapsed time in case of a rollback.<br />

► Use the partitioning of the F- fact table.<br />

– Refer to <strong>SAP</strong> note 860830 (<strong>DB2</strong>_PARTITIONED_F_FACT = X).<br />

– Configure the fast deletion of the partition <strong>for</strong> the request deletion from the<br />

F-Fact able (<strong>SAP</strong> note 894045).<br />

Chapter 6. <strong>Best</strong> practices <strong>for</strong> InfoCubes 101

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

Saved successfully!

Ooh no, something went wrong!