This paper first presents the problem of building a highly available key-value store for Amazon's online shopping platform. It then introduces the Dynamo DB's architecture, which uses a almost all the known concepts of distributed systems like consistent hashing with virtual nodes, vector clocks, hinted handoff, sloppy quorum, etc to make it 'highly available' for read and write and causal consistent. You will get interest once you read as it talks about they've literally used in Amazon's cart service