Out of interest, given that you say you have a 3.5GB blob, 1) How big would you expect an entire database to be ? In other words, how big might the total size of all blobs in the same database file be ? 2) Do you expect to store more than one blob in the same row ? 3) What's your factor of write-to-read on these blobs ? In other words, does the majority of data get stored but go unread ? Or do you expect each blob get read (i.e. named in a <CODE>SELECT</CODE>) many times ? The answers to these questions may help us recommend API calls or techniques which best suit you.