I have a cube that does not run fast enough. I have one dimension that is over 270K members and growing. I am thinking to make a separate cube that includes only a specific 70K or so members and the fact records associated with them. This is easy to do on the measure groups by adjusting the partition queries to include an additional where clause filter. For the dimensions, it seems I would have to create separate dimensions in database including redoing the dsv relationships, dimension design, etc. This is work and also it is cumbersome from a maintenance perspective. Does anyone have a better idea of how to do this?
I feel that this dimension is really slowing down the cube, because there are no more than 5 million records in the largest measure group and the server has fast SAN storage and more memory than it uses, yet still it can be slow depending on the query. I have previously worked to try to optimize it based on query plans and aggregations and so forth without luck. At this point I really don't have time to spend on it, but this seemed like a potential quick solution and it would work ok for the users.
Thanks,
Ken