I've tried changing our process so that all simulations are written to the same table with a unique ID for each simulation.

[SimId] [smallint] NOT NULL,

[ValueIndex] [smallint] NOT NULL,

[TimeIndex] [smallint] NOT NULL,

[Result] [float] NOT NULL

I've then added a clustered index across SimId, ValueIndex and TimeIndex (ValueIndex and TimeIndex are the clustered PK columns on the existing dynamically created tables).

Running a simulation that produces 25 million results takes over twice as long as previously to insert the data.

The speed of reading the data speed is pretty much similar to before, which is good.

Deleting results takes a lot longer as like Kevin mentioned we were just dropping the tables for simulations we wanted to delete whereas now we're deleting a batch of records from the single table.

Darren