Tencent Cloud Database TDSQL Team Achieves Two Papers Accepted at ACM SIGMOD 2021

From June 20-25, 2021, the ACM SIGMOD conference, a top international database conference, was held in Xi'an. Tencent Cloud Database TDSQL team had two papers accepted, demonstrating significant technological innovations.

The first paper, "Spitfire: A Three-Tier Buffer Manager for Volatile and Non-Volatile Memory," proposed the innovative Spitfire concept, a multi-level buffer manager that effectively identifies hot data and keeps it in DRAM cache while storing warm data in non-volatile memory (NVM), significantly improving system performance and cost-effectiveness. SIGMOD reviewers highly praised Spitfire's innovation, noting that its research on combining NVM hardware with database architecture was previously underexplored. This achievement has been applied to TDSQL-C products, improving read performance by over 100% without significantly increasing costs.

The second paper, a collaboration with Renmin University of China titled "Efficiently Supporting Adaptive MultiLevel Serializability Models in Distributed Database Systems," achieved high-performance strong consistency in distributed database systems. This innovation enables distributed databases to achieve "dual consistency" (transaction consistency and distributed consistency) while greatly improving the processing efficiency of distributed transaction clusters.

Previously, TDSQL has been featured at major database conferences including SIGMOD, ICDE, and VLDB, as well as top journals like IEEE TKDE. Notable achievements include a 2019 collaboration with Huazhong University of Science and Technology on "An End-to-End Automatic Cloud Database Tuning System Using Deep Reinforcement Learning," which achieved AI-based database performance optimization results that surpassed expert-level tuning for the first time.

In 2019, Tencent Cloud Database team collaborated with Renmin University to innovatively propose and implement the full-temporal database system T-TDSQL, which provides lightweight full-temporal data management and transaction processing capabilities while maintaining OLTP performance. This system was accepted by VLDB 2019.

TDSQL continues to promote deep integration and innovation between AI, new hardware, cloud computing, and database technologies, investing in basic research, academic-industry collaboration, and talent cultivation to advance the domestic database technology ecosystem.