On June 13, the reporter was informed that three papers from Tencent Cloud Database were again selected into SIGMOD, the top conference in the database industry, and included in SIGMOD 2022 Research Full Paper.
Among the research results included this time, new data structure design, AI intelligent parameter tuning optimization, etc. are all proposed for the first time in the industry. Tencent Cloud Database has been selected into SIGMOD for many times, indicating that Tencent Cloud Database has won international authoritative recognition for its accumulation and cutting-edge innovation in storage, intelligent management and control, etc.
SIGMOD, the full name of the International Conference on Data Management (Special Interest Group on Management Of Data), is an international academic conference initiated by the Association for Computing Machinery (ACM) Data Management Committee (SIGMOD) and has the highest academic status in the database field.
The title of the first selected paper is HUNTER: An Online Cloud Database Hybrid Tuning System for Personalized Requirements, which was completed by Tencent Cloud Database Team and Huazhong University of Science and Technology, and made further breakthroughs in AI intelligent parameter tuning optimization.
Generally speaking, there are hundreds of database parameter settings, and these parameters control the performance of the database. Professional operation and maintenance personnel will spend a lot of time and experience to tune the parameters of the database to match different hardware, requirements and business scenarios. How to use AI technology to solve database system performance problems is becoming more and more important and urgent.
In this paper, the Tencent Cloud Database TDSQL-C team proposed the hybrid tuning system Hunter, which mainly solves the problem of how to significantly reduce the tuning time while ensuring the tuning effect. Experiments show that with the increase of concurrency, the optimization time is reduced quasi-linearly. The optimization time is only 17 hours in a single concurrency scenario, and the optimization time is shortened to 2 hours in a 20 concurrency scenario.
The second selected paper was jointly completed by Renmin University of China and Tencent Cloud Database Team, entitled CompressDB: Enabling Efficient Compressed Data Direct Processing for Various Databases. The paper proposes a new database processing technology for the direct manipulation and processing of compressed data—— CompressDB.
Faced with the exponentially growing amount of data, the industry generally uses data compression to reduce storage space. In a big data management system, directly operating on compressed data can save storage space and improve processing performance. However, current such systems only focus on data query, and a complete big data management system must support data query and data manipulation.
This research proposes and implements a new database technology, which uses context-free grammar to compress data, and realizes parsing of grammar rules through new data structure and algorithm design. CompressDB supports data query and operation directly on compressed data, and supports various kind of database system. Experiments show that CompressDB achieves an average throughput improvement of 40% and latency reduction of 44%, and achieves a compression ratio of 1.81 times.
The third selected paper is a collaboration between Tencent and Peking University, titled BlindFL: Vertical Federated Machine Learning without Peeking into Your Data. For data privacy security issues, this paper proposes a new vertical federated learning paradigm, BlindFL, which can support multiple feature data types and prove its security under the semi-honest security assumption. The experimental results show that BlindFL can effectively protect the private data of the participants and has higher operating efficiency.
Tencent has always attached great importance to the research and development in the field of databases. Through the joint construction of the school-enterprise laboratory, the CCF-Tencent Rhino-Bird Fund for young scholars, and the Tencent Rhino-Bird Scientific Research Project. We will build long-term cooperation with colleges and universities and scientific research institutions to integrate “production, education and research”, convert technological research results into applications, and continue to output cutting-edge technological innovations and demonstration applications through school-enterprise cooperation.
Previously, Tencent Cloud Database TDSQL has repeatedly appeared in the three top database conferences SIGMOD, ICDE, VLDB, and the top journals in the field of data science, IEEE TKDE, etc. Tencent Cloud Database TDSQL will continue to increase investment in basic database research and innovation and the construction of database industry-university-research cooperation ecology, fully release the dividends of leading technologies, and help domestic database academic talent training and technological innovation ecological construction and development.
Leifeng.com
This article is reproduced from: https://www.leiphone.com/category/industrynews/QIeUtlHUBlKDSXe4.html
This site is for inclusion only, and the copyright belongs to the original author.