How about a merkle tree structure for storing data (like Git does)? This would make it easy to find out what the snapshot at any given point of time was. Q is, whether it is powerful enough to support typical data-oriented applications?