Why don't u just load the files into the database using a sql script rather
than thru the application? It's a lot more efficient.

To help performance, you have two options.

1.  Push the entire XML into sql server.  Large overhead, and maybe not what
you want.

2.  (My suggestion).  Instead of pushing the records one by one into the db,
make the number configurable.
For example, lets make it 1000.

Use the IDataReader to loop over the CSV file.
Put 1000 records into the DataSet.
Write a usp (user stored procedure) which accepts @xml_doc text. (which will
be xml).
Do a bulk insert/update using the xml.

You will save alot of time doing this.

There are other options, but based on what you put in your OP, I'd go for
#2.

how do you do a dataset bulk insert/update, you ask?

Go here:
http://groups.google.com/group/microsoft.public.sqlserver.programming/msg/268be0
e152d8b876
read the MS links I provide.

The key is to use the ds.GetXml ....to get 1000 records worth records.
Then push those 1000 records into TSQL as xml.
Parse the xml into a @variable or #temp table.
Do your updates/inserts from there.

my uspProductUpdate is a fine tuned version......

Because index rebuilding happens after the 1000 records get
updated/inserted, you save time that way as well.

..

Then experiment with the 1000 number.  Maybe 5000, maybe 500 is your sweet
spot number.

..