Compression & Generation of Indexes

[Former Member]

Hi,

I am working on compressing some InfoCubes that have never been compressed before. I am creating a process chain that starts the compression job weekly.

It would be a great help if someone can answer these questions -

1. Should I delete and create index before I start the compression step in the process chain? Or is it more beneficial if I do the deletion of index before the compression step and creation of index subsequently after the compression step. I am not sure whether the indexing works on the F table or the E table, that's the reason for this question.
2. In the process chain steps 'Delete index' and 'Create index', I see an option to choose 'Infocube', 'DTP' and "Execute InfoPackage'. I understand the 'InfoCube' part but why are 'DTP' and 'InfoPackage' options here and in which scenarios does it make sense to choose these options.
3. Similary, in the process chain step 'Compression of the InfoCube', I see an option to choose 'Delete Index', 'DTP' and 'Execute InfoPackage'. How are these options useful?

[Former Member · Accepted solution]

Hi,

To answer your questions:-

1-> the indexing is for both F and E tables. we should do deletion of index first to improve loading performance.  The creation of index is for better querying. so before compression,both the steps should be done, as it need querying.

Regarding your 2 and 3rd:- i assume, the DTP and IP are nothing to do here. I tried with DTP option for compression, and i got the message in logs as" there is no path from DTP_LOAD XX to Compress"

so we have to choose the correct object type according to processes.

Regards,
Sakthi.

[Former Member · Accepted solution]

Hi,

Thanks for your reply. Doesn't it make more sense to do the indexing after the compression since the data is summarized (i.e. request id removed) ?

[Former Member · Accepted solution]

Hi Grame,

I was thinking, that compression also needs indexing for better compression performance, because on checking the logs of compression, in the sql, it had used index of tables.

but i checked more now that compression needs the primary index which they call as P index  or something.

So i guess, your process would make it more sense, if compression doesn't need these indexes.

Regards,
Sakthi.

[Former Member · Accepted solution]

Thank you for all the help