I still think merlyn is right -- a db is the way to go. blobs are not the most elegant/efficient mechanisms, but they are very easy to find based on a key. As long as your blobs stay below about 1MB, mysql or postgres should be fine. Trying to find a single file in a directory hierarchy of millions of entries is going to suffer significantly worse performance.