Why are you surprised ? Databases never promised to be fast...

The point of a database is managability of the data and maintainability. Things like ACID properties, remote access, simultanous access, backups, etc. It's that all the sophisticated stuff like memory caching, writeback, indices, logfiles etc. has all already been done for you.

But in the end, there is *no magic*. A database also gets its data from files (or cached memory) and then does a data exchange with you over some form of IPC. For any specific case you can almost certainly handroll a version that's faster because you can skip many steps needed for the powerfull generic version. Even with a relatively slow language like perl it can often be done faster. Even for rather sophisticated setups you can often beat an SQL database by simply using DBM cleverly.

But then, speed was never the point. Avoiding one off solutions for everything is.

(All of which doesn't imply that neither the DBI version nor the pure perl version can't be improved anymore)