硬件事务内存（hardware transactional memory，HTM）能够极大地提升多核内存事务处理的吞吐.然而，为了避免慢速持久化设备对事务吞吐的影响，现有系统以批量的方式提交事务，这使得事务提交有极高的延迟.低时延非易失性内存（non-volatile memory，NVM）的出现，给降低基于HTM的内存事务处理时延带来了机遇；然而，利用NVM需要解决HTM无法和NVM硬件协同的挑战：持久化写入NVM会直接中断HTM的执行.为了解决这一问题，提出了名为Parity Version的机制，将事务中的NVM操作和HTM操作进行区分.这样，事务可以正确且高效地利用NVM降低基于HTM事务处理的时延.将该机制集成到现有基于HTM的内存数据库、DBX中，并提出了DBXN：一个低时延高吞吐的内存数据库.：还针对真实NVM硬件的特性对事务实现进行了相应的优化.在典型事务处理测试基准TPC-C中，DBXN能够将DBX的事务提交时延降低99%，同时还有2.1倍更高的吞吐.

The emergency of hardware transactional memory (HTM) has greatly boosted the transaction processing throughput in in-memory databases. However, the group commit used in in-memory databases, which aims at reducing the impact from slow persistent storage, leads to high transaction commit latency. Non-volatile memory (NVM) opens opportunities for reducing transaction commit latency. However, HTM cannot cooperate with NVM together: flushing data to NVM will always cause HTM to abort. This study proposes a technique called Parity Version to decouple the process of HTM execution and NVM write. Thus, the transactions can correctly and efficiently use NVM to reduce their commit latency with HTM. This technique has been integrated to DBX, a state-of-the-art HTM-based database, and DBXN: A low-latency and high-throughput in-memory transaction processing system, is proposed. Evaluations using typical OLTP workloads, including TPC-C, show that it has 99% lower latency and 2.1 times higher throughput than DBX.

国家重点研发计划（2020YFB2104100）；国家杰出青年科学基金（61925206）