At the 49th VLDB Conference, the top international database conference, two papers of Tencent Cloud TDSQL were successfully included in VLDB 2023, once again proving that innovative technology is recognized by the top international conference VLDB
As a leader in the database field One of the three top conferences, each VLDB conference focuses on displaying the most cutting-edge directions of current database research and the latest applications in the industry, attracting the participation of many of the world's top technology companies and research institutions. Because the conference has extremely high requirements on system innovation, completeness, experimental design, etc., the paper acceptance rate of the VLDB conference is generally low (about 18%).
Among the selected papers, the "Efficient Black-box Checking of Snapshot Isolation in Databases" solution jointly developed by Tencent Cloud, Nanjing University and ETH Zurich proposes a novel black-box checking The processor - PolySI, can efficiently check Snapshot isolation (SI) and provide understandable counterexamples when a violation is detected.
Snapshot isolation is a common weak isolation level, which avoids the performance loss caused by serialization and can prevent many common data anomalies. However, some production cloud databases that claim to provide snapshot isolation guarantees will still generate SI data anomalies, which will have a huge impact, especially in the financial field. Existing similar tools in the industry either do not support snapshot isolation level testing or are less efficient. Given the complexity of database systems and the fact that internal information within the database is often inaccessible, a black-box snapshot isolation checker is urgently needed in the industry.
In order to solve this problem, we proposed and designed the "PolySI" algorithm and tools. The theoretical basis of PolySI is the SI characterization theorem based on Generalized Polygraphs (GPs), which guarantees the correctness and completeness of PolySI. PolySI adopts an SMT solver (MonoSAT) and leverages GPs’ compact constraint encoding scheme along with domain-specific optimizations to accelerate SMT solving
Currently, through extensive evaluation, PolySI successfully reproduces known SI anomalies and New SI anomalies were detected in three production cloud databases, providing understandable counterexamples. PolySI outperforms current state-of-the-art SI black-box checkers under multiple classes of workloads and is able to scale to large-scale workloads.
According to our understanding, the paper "Online Schema Evolution is (Almost) Free for Snapshot Databases" jointly completed by Tencent Cloud and Simon Fraser University introduces a new online and transactional schema evolution called "Tesseract" Method, designed to solve the challenges faced in the process of online database modification of schema
Currently, modern database applications often make schema changes according to changing needs. The main advantage of online database modification of schema is that there is no need to stop the database service or Structural modifications can be made by interrupting ongoing transactions, allowing the database to meet dynamic changes without requiring downtime for maintenance or restarting the database.
In existing database systems, although online and transactional schema (schema) evolution are supported, they also face some challenges. The first is the issue of data consistency. When making structural modifications, in order to ensure the consistency of the data, transactions or other mechanisms need to be used to ensure the integrity and correctness of the data. Secondly, there is the problem of long running time. Some structural modifications may take a long time to complete, especially for large databases or modifications of complex structures, which may have a certain impact on the performance of the database. Therefore, modifications need to be made in an appropriate time window to minimize the impact on the business
In past solutions, an ad hoc approach was often used to "patch" the schema evolution and apply it to the existing system. This resulted in many edge cases and incomplete functionality. Therefore, applications often require carefully scheduled downtime to make schema changes, thus sacrificing availability
To avoid the above shortcomings, Tesseract comes into play. In widely used multi-version database systems, schema evolution can be modeled as data modification operations on the entire table, which is the so-called Data Definition as Modification (DDaM). In this way, Tesseract can support the pattern at almost zero cost by leveraging the concurrency control protocol
In Tesseract application testing, we made a simple adjustment to the existing snapshot isolation protocol. Under workloads on 40-core servers, Tesseract is able to achieve online, transactional schema evolution without downtime, and maintain high application performance during the evolution process
By participating in the top database conference VLDB, Tencent Cloud will The latest technological breakthroughs and innovative directions in the database field are shared with global technology developers simultaneously, which also provides extremely valuable reference cases for technology and industrial development in the database field. In the future, Tencent Cloud will continue to improve database-related technologies, products and ecological capabilities to provide convenient and easy-to-use database services for all walks of life.
The above is the detailed content of Tencent Cloud Database was once again recognized by top conferences, and the paper was successfully selected for VLDB2023. For more information, please follow other related articles on the PHP Chinese website!