cancel
Showing results for 
Search instead for 
Did you mean: 

posting journals very slow

Former Member
0 Kudos

Hi,

when we post a simple journal it takes a long time before the journal has been posted (30 sec - 1 minute):

- we have no journal scripting behind it

- we perform lite optimizes which reduces the throughput time, but the minimum remains on 30 secs for one simple journal.

When performing a trace in BW we notice the system often reads on the facttable of the cube joined with all dimension member tables. This query is executed > 10 times ... and takes up to 0,5 -1 second in average ...

Is 30 seconds normal behaviour. We really think this is very slow ...

regards

Dries

Accepted Solutions (1)

Accepted Solutions (1)

esjewett
Active Contributor
0 Kudos

Hi Dries,

30 seconds seems pretty slow to me. I just tried it on a test system and it took more like 7 seconds.

The behavior where the system reads the fact table joined to the dimension tables is normal for BW. BW is a star-schema ROLAP model, so any query of the cube is going to involve joining the dim-tables with the e/f-tables. This join is fast because the tables are fully indexed on the join key (or at least we hope they are ;-).

The system needs to read the cube whenever you post a journal so that it can write a delta record (the calculated value minus the value already on the database) if you are posting a journal to an intersection that already has data. Journals are additive, but since the journal functionality still goes through the normal write-back routine, I think it still does this delta calculation. I haven't looked at the ABAP for a while, so it could have changed, but that's what I recall.

The fact that the query is running more than 10 times does seem odd to me, but I haven't looked into the query execution for journal posting, so I'm not sure what it it doing. Do the multiple queries run sequentially or in parallel? Is the total time for the multiple queries similar to the total time for posting the journal, or is there something else going on?

Ethan

Edited to clarify about journal delta calculation - journals are additive to values already on the database.

Former Member
0 Kudos

Hi Ethan,

the time for running the queries on the fact table explain most of the processing time. If it would only read it once, we also would come to something like 7 seconds I think ... If you like I can sent you the details of the trace performed in BW (ST05).

Dries

esjewett
Active Contributor
0 Kudos

Hi Dries,

If you can post the details somewhere and link to it, that would be helpful. Maybe in a Google Doc or something like that?

Ethan

Former Member
0 Kudos

Answers (0)