SHAFT: Serializable, Highly Available and Fault Tolerant Concurrency Control in the Cloud

Guaranteeing transaction semantics in a highly available and fault tolerant manner is desirable to application developers. Besides, it is a very valuable feature for database-backed applications. SHAFT is a pessimistic concurrency control protocol for partitioned and replicated data, which can be distributed across multiple datacenters. Laying its basis on the Paxos algorithm, the SHAFT protocol guarantees Serializability, High Availability and Fault Tolerance simultaneously for transactions. The distributed concurrency control process of SHAFT meets the strict two-phase locking requirements. SHAFT can restart a transaction aborted due to the inability to lock its data. High availability is guaranteed as read-only transactions are processed with shorter procedures. Concurrent transactions can be merged together and processed in one SHAFT instance to reduce costs and latency, while increasing concurrency. Different from other existing transactional replication protocols, SHAFT allows a client to actively abort a transaction. SHAFT also allows flexible data partition, replication and distribution, a proper combination of which can reduce costs and improve performance. SHAFT performs well even under failures. Our experiments show that SHAFT outperforms a recent related work MDCC, which outperforms other synchronous transactional replication protocols, e.g.