DynamoDB is a highly available, fully managed, key-value NoSQL database. It is known for its ’always-on’ experience and single-digit millisecond performance at any scale. Based on [1], we examine CAP theorem consequences of offering availability, handling network partitions with a weaker consistency model. We referenced [2] for detailed implementation. We examine the effects of consistent hashing, sloppy and strict quorums in our design. Process failures are handled assuming a fail-stop model (a failed pro- cess never rejoins the cluster, therefore no need of hinted handoff mechanism). View reconciliation in the scenario of multiple node failures (Nodes/processes redistribute key ranges ownership to maintain N -> the replication factor) are tested as well.