I haven't worked much with it, but have you considered using DBD::ADO as an alternative? Reasoning being that ADO is a Microsoft-created data access protocol, and therefore may be somewhat more speedy than using ODBC. And by using DBD::ADO, you could keep all of your existing code, and all future code benefits from what DBI provides.

Give it a shot, you have nothing to lose. Wish I had some benchmarks for you to confirm my theories ;)

Good luck!
MrCromeDome