7 Top-Tier Database Conference Papers Power Core Transaction Speedup of 32% TencentDB TDSQL
SHENZHEN, China, April 24, 2025 /PRNewswire/ -- Renmin University and Tencent's joint innovation lab showcased groundbreaking advancements in financial-grade distributed database systems at their 2025 annual conference. Over two years, seven papers were accepted at top database conferences (VLDB, ICDE, SIGMOD), with core technologies integrated into TencentDB TDSQL, achieving 32% faster transactions in state-owned banks and 40% improved query speeds in government big data applications.
Key Innovations:
1. Transaction Processing Optimization
- HDCC Protocol: Combines deterministic (Calvin) and non-deterministic (OCC) concurrency control, resolving cross-mechanism recovery issues. Delivers 5.7x performance gains (VLDB 2025).
- Lion Protocol: Uses LSTM-based workload prediction to optimize replica placement dynamically, avoiding migration bottlenecks. Achieves 2.7x higher throughput and 76.4% better scalability (ICDE 2024).
2. Storage Optimization
- HOCO Engine: Enables direct computation on compressed text data via homomorphic compression, offering 9.18x higher throughput for access/modification and 7.16x lower latency for analytics (SIGMOD 2024).
- SALI Framework: A scalable learned index that adapts to workload skew and boosts concurrency, delivering 2.04x higher insertion throughput under 64 threads (SIGMOD 2024).
Collaboration Impact: The research bridges academia and industry, with contributions from Renmin University's data engineering experts and Tencent's TDSQL R&D team. Attendees included leaders such as Du Xiaoyong (Renmin) and Wang Juhong (Tencent), highlighting the partnership's strategic importance in advancing database technologies for critical sectors like finance and governance.
This collaboration underscores China's growing leadership in high-performance database systems, combining cutting-edge AI/ML techniques with robust engineering to solve real-world scalability and efficiency challenges.