thx for the fast answer!
the reason I did not post "source data structure" is that I wanted to see if Perl has solutions...
The data field I am concerned about is a mysql "text" field called "description" and may contain long text. The other is just the according id. The rest is index and fetched via sphinx.
And regarding speed: never fast enough :) as "first come first serve" is the fixed sorting order and we have to be faster than the others :)
Comment on Re^2: read a huge mysql Table into memory
Well, that's pretty simple table and I'm sure you have index on id column, so 200 queries per second shouldn't be a problem for mysql if you just selecting records by id. Anyway as you already saw perl requires too much memory to store your data, so it seems you're tied with database solution.