I don’t know where this topic exactly belongs but the use case is as follows:
We have a big fact table and a dimension. The fact table changes once per day but the dimension table is modified by the application constantly (user input). The queries run of course on both with aggregate and filtering functions. The user’s expectations are that changes in the dimension table are reflected after short time (<5s) in the query results.
So invalidating or a forced partial pre-aggregation should be kicked off in an evented manner (we know what to preaggregate and when). Is there any recipe how to do that? With cube store? Any numbers/experiences how big the delay is? costs?
Thx for any input