Efficient Way Of Backing Up Data

We are writing an in-house backup tools. It's written in C# and the program allows me to select which records we want to archive and remove from existing database. Functionally, we can acheive what we want to do but I feel that we did not do it right. Can someone show me the right way to go? Here is what we are doing in the backup process.

1) Create an archive database (with the file name, say archive_101.mdf)
2) Move the data from orginal database to archive database using the following SQL statement
INSERT INTO [archive_database].dbo.records SELECT a, b, c, date FROM [original_database] WHERE date < CONVERT(DATETIME, '2007-10-10', 120)
3) Delete the moved data from original database.
4) Of course, step 2 and step3 are done in a transaction
5) Then, we detach the archive database and store the file "archive_101.mdf somewhere.

I felt it's wrong because the performance is very bad. If I run the select statement, it takes me 1 min to dump all the data to console. If I run the same SELECT statement together with INSERT INTO, like what I wrote in step 2, it takes me more than 1 hour to write to another database. I checked the tempdb and its size does grow a lot when I am doing step 2. I know the data I am selecting is about 500MB large but still, it should not take that long. Can somebody give me any hints?

Thanks in advance.