You wouldn't start with Pg but it can be a destination. DBD::Pg suffers by having to read the entire result set into memory before using it. Pg can perform both asynchronous queries and has support for cursors. If you're dealing with large data sets then DBD::Pg may just be entirely inappropriate. There's a Pg::Smart or something named like that that picks up where Pg left off and is more DBI-like except with all the good bits.