Better practices for improved APD performance.

APD plays a major role in data mining which enables to use queries, database tables, and files as well as Info-Providers as data sources in the analysis process to perform ad-hoc analyses and the data can stored in the BI system (Data Store objects for direct update or Info Objects with attributes) or in a CRM system.

The following authorisation objects are required before we proceed with creation of APD.

1) While using  APD  for Segmentation purpose, please click on the data target option which can be seen in RSANWB  T-code  . After choosing the target option , you can  you user ID or  Admin User ID will be checked for C_CRMBWTGT authorisation . once its successful , you will be able to define the target destination  along with the logical system , sub object. Please note the process to update CRM table will be running only in Dialog process which might take some time . Key pre-requisites ,

Source is always preferred for  Direct Update DSO in order to avoid inconsistency .

Avoid writing Massive ABAP transformation Logic. If it is required , then create an intermediate DSO before updating it to CRM which could provide better performance.

2)  While choosing the source for APD , you could notice 2 options  Read attributes of a characteristic & Read data from the Info provider . If you want all the columns from a master data info provide you could prefer the first option else go with next one with which you can choose the required fields using the field selection tab . In doing so , the memory required to perform join would be minimised and APD will give better performance.

3) If a particular info object has been created as  display only attribute , they wont be visible for further transformation of data. In that case , we need to chose the source option ” Read data from the Data base table” , which would allow to extract the display only fields as well.

4) While choosing the Query as an Info provider ,  The Data package size and partitioning characteristics are preferred to be maintained so as to yield faster execution of APD.

Now lets have an idea on some of the frequent issues  and the steps to resolve those.

We have faced the following issues in different scenarios.

1)  ORA-01652: unable to extend temp error

Reason:  There are 2 possible reasons for this error. A) The PSAP Temp table space memory is not sufficient enough to process the mass volume of data & b) if the APD is reading data from the Info provider which has BWA indexes built on it.

Solution: the parameter RSAN_APD_DS_INFOPROV_NO_BIA in table RSADMIN should be set to X which will skip the APD hitting the BWA, which doesn’t allow more than 200,000 records while reading data.

2) Deadlock while running APD with error message X.

Reason: The APD does created internal temporary tables based on the number of joins in it. After the jobs get completed successfully, the internal tables will be deleted. Sometimes there may be some delay in doing so and if we start the ADP immediately, the deadlock issue will happen.

Solution:  AS per the SAP note 1715230 , either we need to delete the table using the program or need to advice the DB team to delete it directly from the Data base.

Hope this will helps in order to understand the issue .