It really depends on what you're doing with the data. Whether you're simply reading in everything from the file and using everything (as opposed to reading in an entire flat file and then only using 5% of the data) or whether you're constantly goig to the DB an finding indexed items, etc. There are so many factors relating to how efficient one way or another performs that, perhaps, the best way to know for your specific application (and DB setup - connection pooling, indexing, caching, etc.) is to write some
test implementions using both methods and checking the speed for a connection, performance under heavy load (if that could be an issue), etc.. Hope this helps.
Sean