Efficient Locking Techniques for Databases on Modern Hardware

Traditional database systems are driven by the assumption that disk I/O is the primary bottleneck, overshadowing all other costs. However, future database systems will be dominated by many-core processors, large main memory, and low-latency semiconductor mass storage. In the increasingly common case that the working data set fits in memory or low-latency storage, new bottlenecks emerge: locking, latching, logging, and critical sections in the buffer manager. Prior work has addressed two of these – latching and logging. This paper addresses locking and proposes new mechanisms optimized for modern hardware. We devised new algorithms and methods to improve all components of database locking, including key range locking, intent locks, detection and recovery from deadlocks, and early lock release. Most of the techniques are easily applicable to other database systems. Experiments with Shore-MT, the transaction processing engine we used as the implementation basis, show throughput improvement by factors of 5 to 50.