How can a variety of unstructured data be stored in a scalable manner such that it can be randomly accessed based on a unique identifier?