Physical Partitioing Calender Year
Cube is physically partitioned on 0Calmonth.
Report is running on Calender Year Example - 2007 and 0calmonth2 Example - 005 not on 0Calmonth Example - 2007005.
Does phy part still improve performance.
Message was edited by:
Chetan Patel replied
Table partitions are physically separated tables, but logically they are linked to one table name. PSA tables and non-compressed F-fact table are partitioned by the system (by request ID). The (compressed) E-fact table can be partitioned by the user by certain time characteristics. For range-partitioned InfoCubes, the SID of the chosen time characteristic is added to both fact tables.
When using range partitioning, query response time is generally improved by partition pruning on the E fact table: all irrelevant partitions are discarded and the data volume to be read is reduced by the time restriction of the query. Note that all F fact table partitions have to be read.
The number of (range) partitions is one more than defined. This additional partition keeps all time characteristics higher than the upper limit. Example: monthly partitions 01/2007 12/2007. This creates 12+1 partitions. The first one contains records from 01/2007 and earlier, then there are 11 partitions containing the months 02 12/2007. An additional partition for all records after 12/2004 is created.
In ORACLE, report SAP_DROP_EMPTY_FPARTITIONS can help you to remove unused or empty partitions of InfoCube or aggregate fact tables. Unused or empty partitions can emerge in case of selective deletion or aborted compression and may effect query performance as all F fact table partitions are accessed for queries on the InfoCube.
Hope it Helps