Two-phase commit (2PC) has been widely used in distributed databases to ensure atomicity for distributed transactions. However, 2PC suffers from two limitations. First, 2PC incurs long latency as it requires two logging operations on the critical path. Second, when a coordinator fails, a participant may be blocked waiting for the coordinators decision, leading to indefinitely long latency and low throughput. We make a key observation that modern cloud databases feature a storage disaggregation architecture, which allows a transactions final decision to not rely on the central coordinator. We propose Cornus, a one-phase commit (1PC) protocol specifically designed for this architecture. Cornus can solve the two problems mentioned above by leveraging the fact that all compute nodes are able to access and modify the log data on any storage node. We present Cornus in detail, formally prove its correctness, develop certain optimization techniques, and evaluate against 2PC on YCSB and TPC-C workloads. The results show that Cornus can achieve 1.5x speedup in latency.
Download