You can query by arbitrary dimension (what we would think of as time aggregates). In the future, you'd be able to materialize to arbitrary dimensions with continuous queries.
How do you support arbitrary dimensions in a tabular database system without massively blowing up the number of tables, the number of columns, or number of sparse rows?
This kind of problem is really an OLAP cube problem, yet you're throwing an OLTP solution at it. I've seen it tried before at several sites and it's been abandoned each time for having crippling performance and management behaviors. What are you doing differently that will make this attempt successful?
