BW process chain flow - optimal performance for cubes

[SvenS]

Hello

I have a question towards the performance of process chain flows (specific the flow for loading in data for cubes in BW). I'm currently working on a new project towards process chains. I'm now looking into the old flows that already exist and I see 2 differences concerning the data load of cubes.

Flow one that I came across was as followed (this is also the way it was done most of the time): Drop index -> load data -> create index -> DB statistics -> compression.

Second one I came across is: Drop index -> load data -> DB statistics -> compression -> create index.

Could someone explain to me what the difference would be between these two in performance and why it would make a difference, if there is a difference at all.

Kind regards
Sven Swennen

[john_hawk · Accepted solution]

Hi Sven,

In terms of query output performance, the sequence of actions does not matter.

In terms of load performance, there may be some slight advantage to executing compression first, since the compressed InfoCube may have slightly fewer rows. If you have multiple processes running in parallel, there might be an advantage in avoiding multiple concurrent compressions since compression are the most resource intensive operations.

Ultimately, either path is acceptable.

John Hawk