Evaluation Review | GBase 8c Distributed Transactional Database
The GBase 8c distributed transactional database passed the 13th batch of CAICT distributed transactional database performance evaluation with outstanding performance and high availability. Its distributed transaction performance, ACID capability, and high availability for financial core business all met the evaluation requirements.
As a financial-grade distributed transactional database, GBase 8c delivers high performance, high availability, elastic scalability, and strong security. It can be deployed on physical machines, virtual machines, containers, private clouds, and public clouds, providing secure, stable, and reliable data storage and management for financial core systems, internet business systems, and government/enterprise systems.
Adaptive Transaction Processing
GBase 8c uses an adaptive transaction processing mechanism to boost performance. For transactions that can be handled locally on a single node, the coordinator processes them using the local transaction flow without two-phase commit, improving efficiency. For cross-node transactions, the coordinator orchestrates two-phase commit among participants to ensure global consistency. The entire transaction process is transparent to the client.
Parallel Processing
GBase 8c leverages parallel processing technology to enhance performance and throughput:
1. The coordinator generates distributed execution plans, pushing operators down to data nodes for parallel processing.
2. Each data node uses a multi-threaded architecture for concurrent execution.
3. Multi-version concurrency control (MVCC) resolves read-write conflicts, enhancing parallel read/write processing.
Global Deadlock Resolution
GBase 8c supports global deadlock resolution. A detection process collects and analyzes lock-wait information across the cluster. When a deadlock is found, the transaction with the highest global transaction ID is rolled back to resolve the deadlock, significantly improving the performance of simple queries, inserts, deletes, and updates under high concurrency.
Specifically, GBase 8c deploys a global deadlock process on CNs and DNs that communicates with the GTM. It periodically detects transaction wait relationships and sends the results to the GTM to build a global wait-for graph. When a deadlock cycle is detected, the transaction with the largest transaction ID is immediately aborted to solve the global deadlock.
Transaction State Persistence
GBase 8c preserves transaction state. If a coordinator node (CN) fails, active transactions on that node remain intact and can continue running on another CN that takes over, ensuring peak performance with high availability.
Since 2015, the China Academy of Information and Communications Technology (CAICT), as a national high-end professional think tank and platform for industrial innovation, has progressively carried out database technology research, standards development, benchmarking, policy support, and ecosystem building. Over the past six years, CAICT has witnessed continuous improvements in top-level design for China's database industry, breakthroughs in key technologies, a thriving industrial ecosystem, expanding application areas, and well-organized standardization efforts. It has become the country’s most important supporting unit in the database field and has built the most authoritative third-party database evaluation system in China. The product evaluation system covers basic capabilities, performance, stability, and security; the service assessment system includes general service capabilities and specialized service capability evaluations. Evaluations are conducted every six months. As of the end of November 2021, 13 batches of product capability evaluations had been completed, involving 58 enterprises, 93 basic capability tests, 35 performance tests, and 4 stability tests.
Basic Capability Tests have been launched for distributed analytical databases, distributed transactional databases, relational cloud databases, time-series databases, graph databases, key-value in-memory databases, and document databases.
Performance Testing: The CAICT Cloud Computing and Big Data Research Institute has independently developed multiple performance benchmarking tools and adapted them with mainstream products, including the time-series database performance benchmarking tool DataBench-TS for connected vehicle scenarios, the transactional database performance benchmarking tool DataBench-T (open-sourced) for financial core system scenarios, and the analytical database performance benchmarking tool DataBench-A for core financial and telecom scenarios. Compared with TPC-series benchmarks, these tools are better aligned with China’s actual application scenarios. They can execute tests in a graphical, automated manner on uniform hardware provided by CAICT, delivering fairer and more objective comparisons of database products.
Stability Testing: China’s first database stability testing tool based on chaos engineering has been launched. It injects disturbances such as CPU, memory, disk, network, and thread stress to observe the resilience of a database system. This evaluation system has witnessed the maturation of China’s database products and the diversification of the product ecosystem, defining the first tier of domestic database vendors and serving as a bellwether for database industry development.
In terms of database service capabilities, three batches of assessments have been completed, covering 14 projects from 8 enterprises.
Service capabilities are divided into general service capabilities and specialized service capabilities. General service capabilities follow the “Database Service Capability Maturity Model” standard and are assessed for service providers across three capability domains: planning and design, implementation and deployment, and operation and maintenance, subdivided into 27 capability items. Each domain can be evaluated independently. Assessment results are graded from low to high: Initial, Repeatable, Robust, Quantitatively Managed, and Optimizing.
Specialized service capabilities follow standards such as “Classification Requirements for Database Application Migration Service Capability” and “Classification Requirements for Big Data SQL Quality Management Platform Capability,” evaluating common toolchains used in heterogeneous database migration, SQL quality management, etc. Results are graded Basic, Enhanced, and Optimized. The first batch of database application migration evaluations has been completed, with Alibaba Cloud, Huawei Cloud, and Tencent Cloud rated as Optimized, Optimized, and Enhanced, respectively.
First Batch of SQL Quality Management Platform Evaluations Is Now Open
14th Batch of Trusted Database Evaluations Is Underway
In 2022, CAICT upgraded its database evaluation system to the “Trusted Database” evaluation system. Registration for the 14th batch of Trusted Database evaluations is now open. This batch will be completed by mid-May, and after a centralized review meeting to assess the results, certificates will be issued and results announced at the June Big Data Industry Summit.
About the Database Application Innovation Lab
Database software is a fundamental component of the IT infrastructure and a core technology driving the global digital economy. To elevate China’s database industry from a follower to a leader, the CAICT Cloud Computing and Big Data Research Institute established the non-profit, free community Database Application Innovation Lab (the “Lab”) in October 2019, together with numerous database vendors, end users, system integrators, database service providers, and hardware manufacturers, forming a “joint fleet” for China’s database industry. The Lab is dedicated to promoting innovation and development of China’s database industry. It addresses real-world problems with a win-win approach, integrating government, industry, academia, research, and application forces to collaboratively advance database application innovation. The Lab remains an open community and continues to welcome applications from enterprises, institutions, and organizations in the database field.
Lab Member Organizations