Ah no. ADO or ADO.NET were never intended to be bulk operation
interfaces--not until ADO.NET 2.0 which introduced access to the SQL Server
bulk copy interfaces. Check out the SqlBulkCopy class. It can import
flat-file data far, far faster than any query-based program. See
http://www.developer.com/db/article.php/3702826 or my book for more details.

If you want speed, use a Process object to encapsulate BCP, the Bulk Copy
bits from SQL Server. As long as the file is set up correctly, or you have a
mapping file, you will find that it uploads very quickly. There is also
BULK_INSERT, if you need the additional features.

We successfully used this method to load files with millions of records (up
to 25 GB in size) into SQL Server, so I know it works.

One caveat if you are using IDENTITY on the table. If the system is actively
being added to, you should count the records first and then use the DBCC
command to increment the IDENTITY. You will then have to add identity to
each record (ripping through file with a FileReader/FileWriter?) prior to
uploading.

Another option is load the bits you need into the DataSet, as you are, and
then send the XML to a sproc that uses the XML features to insert records.
You basically end up using the DataSet XML as a "table" in the sproc. This
is slower than BCP, but works well when you cannot map.

*************************************************