We have a pretty large cube, with thousands of partitions. The way pour process works is that if there is a cube schema change, like adding new dimension or removing one, etc..., then we drop the SSAS DB and regenerate the cube from scratch in an automated process through AMO calling XMLA scripts.
The problem is it is now taking 7+ just to generate an empty cube shell, that is, a fully structured cube with all measure group partitions, but no data. Data comes later.
We do partitions by day, so every new day there is a new cube partition per MG.
Our data goes back to 2009. So 4 years x 365 partitions x 13 measure groups, that 18,980 measure group partitions that have to be generated, and it's growing with each passing day. Our data is based on daily usage patterns, this is why we have partitioned by day.
Aside from beefing up hardware, are there any directions we could go in or suggestions for making the regeneration of an empty cube shell speed up? I'm almost certain that the slow down is in generating the large number of partitions.
Wondering if perhaps partition creation can be palatalized in xmla, haven't found anything like that, other than for processing.
Thanks!