If you decide to use aggregates you can let BW design aggregates for you, in which caseyou get many aggregates based on navigations in statistics, or design a few levelsyourself, which means you control the extra storage. I tend to create a summary levelaggregate, and mid level aggregate and sometimes additional customer without materialor material without customer aggregates.As you are using Oracle did you also know you can run an 'explain' on the SQL statementof a poorly performing query - run the query -> trans SM66 to find the Process ID (PID)of a the query -> trans ST04 -> detailed analysis -> oracle session -> select SQL textfrom DBA_VIEWS where PID -> use DBA tools to run an explain on the SQL statement.From this you may also find some areas for improvement. Not much to add to the above great answer: looks like you're a textbook example to useaggregates.In addition also have a look at compression: note 590370 explains the advantages youcould get from it. The above note may mention this , but in order to reap the benefits of fact table partioning you need to compress your infocube.Database partioning works on the 'E' table within the infocube (not the 'F' fact table thathold the uncompressed data).
How to build an extractor?
Using SAP BW 2.1C and soon 3.0A we have to extract data from R/3 FI.
Standard extractor can´t be used, since data need to be enhanced and go througvarious routines before data contains all information needed in BW. Delta uploadmechanism is needed.
Where can information on
- building a new extractor (including delta mechanism)
- enhancing a standard extractor be found?
Technically, programming a function module data source would be a modification.There's also a way that allows you to stay close to the standard, especially if you extractdata from standard tables:In OLTP:1) Find out which change document object points to the table.2) Use the generic extractor to build a data source for _master data_ over the table.3) Use the "enhance extract structure" feature in the BW img to add computed fields tothe data source4) Use the user exit for extraction to program your enhancement logic and computeadditional fields