Distributed Hash Tables (DHTs) were originated from the design of structured Peer-To-Peer (P2P) systems. A DHT provides a key-based lookup service similar to a hash table. In this paper, the authors present the detailed design of a new DHT protocol, Tambour. The novelty of the protocol is that it uses parallel lookup to reduce retrive latency and bounds communication overhead to a dynamically adjusted routing table. Tambour estimates the probabilities of routing entries' liveness based on statistics of node lifetime history and evicts dead entries after lookup failures. When the network is unstable, more routing entries will be evicted in a given period of time and the routing tables will be getting smaller which minimize the number of timeouts for later lookup requests.