In clustered file storage, what does the system do with records that have similar characteristics?