I have a cube which has 20 dimensions. There is a dimension for account id's on which I'm doing category count. I don't actually need this account dimension in my cube, but just to have ceategory counts I'm using it. There are more than 4 million categories in it, hence the build time is more than 10 hrs. and cube size is 2.5 GB.
Is there any alternate solution to get the result similar to what I'm acheivng through category count, across all dimensions.
Also, is there any way to speed-up the cube generation process, I have tried creating separate datasources in the transformer,and it has given me a performance improvement of 18%, but to have join between the two datasources I have included the common column of the two datasources, as a level in one of the related dimesnsion and in turn it is increasing the size further by 15%.
I'm using cognos 8.4 and the database is oracle 9i
I had the same issue about a month ago and the only way I could see was to do an external roll-up for that measure (account count). I know it will imply the creation of every single level of detail/aggregation but I did it with views and, well, it took me a few hours to set it up but now it is up and running much faster.
Regards,
Bark
Thanks for the suggestion Bark.
I'm using report as a datasource for this cube. I'm not exactly sure how to create a external rollup for the account count. Creating every single level of detail/aggregation - should I do this for all 20 dimensions. Could you please explain a bit more on this.
Not only for the 20 dimensios but for all of their intersections. Have a look here:
Example: http://www-01.ibm.com/support/docview.wss?uid=swg21349247
How To: http://publib.boulder.ibm.com/infocenter/c8bi/v8r4m0/index.jsp?topic=/com.ibm.swg.im.cognos.ug_cogtr.8.4.0.doc/ug_cogtr_id6364CreateCubeswithExternalRollups.html
Regards,
Bark