Keith,
Thanks for your questions.

1. What kind of box is SQL Server running on compared to VFP? Please give the CUP/RAM/Storage subsystem for both.

2.4 GHz, 1GB RAM, SCSI Raid 5 drives server holds both SQL and VFP tables.

2. Is the processing of these large datasets occurring on the server (Stored Procedures), or on the client?

Stored Procedure.

3. Have you considered importing the data directly from the text files into SQL Server tables using the ODBC text driver?

It's not the text to table that is taking the time, it's the processing of each text line. Basically we're rating each line of information based on several lookups into underlying rate tables. Due to defaults (if a rate record exists for an over-ride field value, use it; else use the record for a default field value) and best case matches we have to use four sequential SQL reads to get to the rate record actually required. All of this is in a single Stored Procedure.

4. Does the project require a large amount of concurrent access?

No - a single PC is dedicated to processing all files as the come in.

5. Are the tables basically flat (non-normalized)?

The rate tables are normalized

6. Can the client handle asynchronous data access?
7. Is the data purged from the database after the processing is complete?

For the tests, yes. In actual processing we start new tables every month.

8. Do the tables contain primary keys and indexes?

Yes.