See discussion Flat File Database, and dbm.

I have used Flat Files for enormous "text" data storage in fixed-length records, and with the indexing to those records done by using binary SDBM files of key/value pairs tied to Perl program Hash tables. The "value" is the record offset to seek to for random access. I normally have about 5 million records in each Flat File, and in each SDBM file where the Flat File is just under 4GIG and the SDBM file about 1 GIG in size.