Personally, unless datafile2 is static, or you can arrange to have whatever is writing that file write directly to the database, I think that using RDBMS is likely to involve more trouble than its worth. The code involved to normalise and insert 80MB+ into the RDBMS is likely to take longer to write and longer to run than processing the data with perl. If the datafile is static that maybe mitigatable, but if you have to reprocess or keep munging large volumes of flat file in order to put it into the database, then the overhead doesn't seem worth it to me.

Further, performing a join on tables where one of them is that size and every record has to be joined with one or more records from the smaller table is going to need quite substantial box and is not going to be quick.

I would be of a different opinion if you can arrange to write the bigfile data directly to the RDBMS and you need to perform one or more additional processings of the big file data in some relational manner on a regular basis, then the process of conversion would start to make sense.

As the popular disclaimer goes, just my 2 cents.