I'd suggest that the size of the dataset is not the deciding factor bewtween using something like Text::CSV_XS and DBD::CSV, although it could be a factor in deciding between DBD::CSV and more robust DBMSs like SQLite or PostgreSQL or MySQL. IMNSHO, the deciding factors in choosing between a Perl parsing approach and a database approach would be the comfort level of the programmer with SQL and the kind and complexities of operations. Sure, parsing modules and parsing without modules can delete, and insert, and update, and sum, and average, and modify column structure, etc. But when one finds oneself doing many of those operations, a database approach may be called for.