cancel
Showing results for 
Search instead for 
Did you mean: 

InfoCube Load Performance

Former Member
0 Kudos

Hello Experts,

I am a beginner in SAP-BI. While loading data from DSo to Infocube i ran into few problems...

Here is the scenario:

- I am loading data into two Info Cubes called ZXYZ, ZABC and my source for this is a DSO(ZDSO) which is same for both the Info cubes.

- However when i triggered the both DTP's at same time, DTP to load into ZABC( ZDSO>ZABC) is running very fast ,Its taking 1-2 Mins to load 50000 records, but the DTP to load ZXYZ (ZDSO>ZXYZ) is running very slow and its taking 30 -45 Mins to load the 50000 records.

- My question is how to speed up this load? Also when you load the data into Info Cube will there be any Internal table comes into picture? If so what about the indexing to the table...

Please help me in speed up the load. Moreover there were no custom routines on the transformations on both the Infocubes.

Thanks,

Srikanth

Accepted Solutions (0)

Answers (4)

Answers (4)

former_member188080
Active Contributor
0 Kudos

Hi,

It depends on many situations

1. You will have transformation..one cube may have direct 1:1 mapping and another may have complex routine and logic maintained in the transformation..so it might take time to load data through this complex transformation..you can debug and check as well on this where exactly it is taking more time...you can modify logic to certain extent..

2. check data packet setting RSCUSTV6

3. Check on semantic group..

4. check whether any summary level or line item level data load that is happening..this also might make difference..

Thanks and regards

Former Member
0 Kudos

Hello,

What i understand is because you are triggering the loads into two cubes at the same time, the DSO table might be accessed by two jobs at the same time which might be resulting in a lock.

1. Check for possible entries in SM12.

2.As others suggested make sure you have deleted indexes on both these cubes before loading.

3. Constantly check SM50 which will show you the tables being accesed.

4 If any of these show NRIV table, buffer the dimensions accordingly in SNRO and then load data.

http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/c0a6e3ff-f000-2d10-b59d-8ef35e248...

Hope it Helps!

Chaitanya

Former Member
0 Kudos

Hi,

You can do the following things

1. Deletion of indexes before load and create the indexes after completion of both the loads.

2. Check whether any possibility to tune the ABAP code in the transformation of long running load.

3. Check the DTP setting ( removal of semantic group, number of parallel process, put more filters on keys, removal of error handling mech )

Regards,

Siva

Former Member
0 Kudos

Hi

Are you deleting the indexes of both the info cubes before you are loading.

how many parallel jobs you have for each DTP. check in DTP --> menu bar, settings --> settings for batch manager.

display the details in DTP monitor, you will come to know which steps is taking long time.

Regards,

Venkatesh

Former Member
0 Kudos

Hi,

As of now we are not deleting the indexes before the data load. This is a test load and we are just testing it. We have all those settings like Delete Indexes & Rebulid Indes in Process Chains. But we are running the DTP's manullly ...will this be an issue...

Thanks,

Srikanth

Former Member
0 Kudos

Not a problem unless you are sure which request you are loading for test loads. You can schedule these loads & wait till the compelete, take a note of the time taken individually for both these loads. I beleive this issue could be because of start/end routines or wrong dimension linkages in the 2nd infoprovider. Anways, deleting indexes before loads is a best practice. Thanks!

former_member186082
Active Contributor
0 Kudos

Hi,

Have you checked the transformations? For the second cube, records converted may be more, hence may take more time for load.

We have one more possibility of locks as both the DTPs access the same ODS. Try running the DTPs one after the other and test the results.

Deleting indices before load is always suggested

Regards

Chandu