Quorum-Based Optimistic Concurrency Control in Replicated DHTs

Distributed hash tables (DHTs) provide a scalable solution for data sharing in P2P systems. To achieve fault tolerance and ensure high data availability, DHTs typically rely on data replication. However, nodes churn, lacking of global view and existence of malicious nodes in P2P environment bring new challenges to the concurrency control for replicated data. Agreement-based algorithms require all-to-all communication among nodes, which does not scale well in P2P environment. Pessimistic approaches perform poorly when clients are malicious. Regards to replicated DHTs, an efficient quorum-based optimistic concurrency control algorithm is proposed.Hybrid failure model is used to reduce the fault-tolerant cost. The responsibilities of data storage and concurrency control are separated to enable data reliability and data availability to be adjusted independently. Simulation shows the proposed algorithm can obviously improve system scalability and reduce the fault-tolerant cost.