in reply to Querying a Hash like a database

A query per second on average should be ridiculously easy with a data set of only 50,000 records. Are you using indexes for the high volume, low return selects?