Key Consistency in DHTs

Most DHTs are designed more for scalability than for consistency, and thus do not provide strong guarantees on the consistency of data. In this paper, we focus on key consistency rather than data consistency: key consistency requires that no key be owned by more than one root. We briefly show how key consistency can be used to support atomic DHT operations and then propose a mechanism to achieve keyconsistency. We have tested our algorithm through simulation and a Planetlab deployment, and find that it provides high availability in the face of node churn and packet drops.