The DMVs I've looked into have the CPU usage with only millisecond accuracy, which is fine for a large enough time window, but not at a batch level.  It looks...

To explain...

Each database is used for a different customer.  My software processes stuff for each customer.  If there's excessive CPU usage, that bumps up their invoice, based (among other things)...

My goal is to know how much CPU time is used on a per-database, per-time-window basis.  I'm not comparing anything, although I may decide to move a database to a...

Hi Michael.  Without going into my reason, I need to know the sum of CPU usage for ad-hoc batches of SQL statements I execute, on a per-database and per-time-window basis. ...

Hi Jeff.  Thanks, but unfortunately, the CPU usage in sys.dm_exec_requests and sys.dm_exec_query_stats are only accurate to the millisecond, and I need more accuracy, since otherwise the result will almost always...

By "batch", I mean one or more statements.  I want to know if I can add any SQL statements to the batch, send that new batch to the SQL Server,...

Hi Gail,

Yeah, I only call them "nested transactions". I'm doing the single transaction thing with SAVE TRANSACTIONs in the called sprocs (depending on whether @@TRANCOUNT is 0), but thanks...

Hi Eirikur,

I don't think you understood my reply. When there are nested sprocs, writing the table variable to the log file can only occur by the outer-most calling sproc,...

I don't think either of those solutions would work, because if a calling sproc were to rollback, any write from the table variable to the log table would be rolled...

That's an interesting approach, although when dealing with nested sprocs and nested transactions, can get a little complex, since AFAIK, you can only pass a table variable as readonly.