One note here. The OP did not mention the size of the CSV data file. DBD::CSV uses Text::CSV_XS under the hood, but it will have to read the complete file into memory to be able to do any database-like operations. With a file of 2Gb, that might result in say 20Gb of memory use (perl overhead). When files are that big - again, I don't know how large the file of the OP is - switching to basic streamed IO processing is usually a lot easier.

I fully agree though that DBD::CSV is the best step towards RDBMS's where those memory limits are not applicable (for the end-user script).

YMMV

update: I just did a quick test with the OP data extended to a 1Mb CSV file. Reading that into memory using getline_all () resulted in a 10Mb data structure (reported by Devel::Size::total_size ()).