Which Amazon storage do you think is the best for my database-style applications that frequently encounter many random reads and writes across the dataset?