Galaxybase breaks Linked Data Benchmark Council (LDBC) Social Network Benchmark (SNB) audit test record with top-notch performance

HANGZHOU, China, May 31, 2022 /PRNewswire/ — About May 16, 2022, LDBC-SNB has released the latest audit test performance record set by Galaxybase, a high-performance distributed graph database developed by CreateLink Technology. Galaxybase saw a massive 70% improvement in throughput and over 600% increase in average query performance over the previous record, as well as rigorous testing on system stability, uptime, accuracy of results, transactional support and recoverability.

The test was performed by an independent and impartial third-party auditor appointed by LDBC on a standard cloud environment. The preparation of the test environment, the generation of the data, the loading of the data, the execution of the tests and the verification of the accuracy of the test results were all strictly in accordance with the LDBC-SNB specifications. To further ensure the credibility and validity of the test results, LDBC conducted a detailed audit of the test code, the entire test environment, and the test process. A full disclosure report is publicly available for To download on the LDBC home page.

Dr. Chen ZHANG, Founder and CEO of CreateLink Technology, will deliver a keynote speech titled “New LDBC SNB benchmark by Galaxybase: more than 6 times faster and 70% more throughput” to share the details of the audit during the ACM SIGMOD/PODS 2022 International Conference on data management on June 17 2022, 11 a.m. EDT, 11:00 p.m. BST. (click to view online)

About the LDBC and SNB test

The Linked Data Benchmark Council (LDBC) is a globally recognized organization for benchmarking graph database standards and audit testing. It brings together the world’s leading industrial hardware and software giants such as Intel, Oracle, as well as experts and scholars from institutions of higher education. LDBC aims to develop fair, honest and comparable methods and mechanisms for measuring graph database management systems and to promote the development of this advanced technology throughout the world, in collaboration with its member organizations.

Social Network Benchmark (SNB) is one of the benchmark suites developed by LDBC. The test suite consists of two separate workloads on a common dataset: the interactive workload and the Business Intelligence (BI) workload, corresponding to sets of tests that respectively measure the performance of data from chart in interactive queries and business intelligence queries. Compared to stand-alone test cases with simple metrics, LDBC-SNB is not only more similar with complex real-world business query scenarios, but also places higher demands on concurrent execution and transaction processing capability. graph database systems.

About Test Details

The audit test performed was the LDBC_SNB interactive workload. Galaxybase has successfully passed the audit test with verification of system configuration compliance with the reference description and its stringent requirements regarding accuracy of results, transaction support, recoverability of the system, zero latency, high throughput and low response time, etc. In particular, Galaxybase ACID test researched serializable isolation level, which is stricter than the committed read isolation level required by the SNB Interactive specification. Additionally, Galaxybase passed the recovery and durability validation test, in which the system was shut down and restarted when the continuous benchmark test run reached 2 hours, and the data from the last record successfully inserted into the LDBC log remained intact and durable in the graph database.

During the performance test, Galaxybase used 48 clients to send concurrent requests to stress test the system. The result showed zero waiting time which is well within LDBC-SNB’s timeout requirement of no more than 5%. With all official audit requirements met, Galaxybase surpassed LDBC’s previous record (held by TuGraph) for all three datasets, namely 30G (80 million vertices, 500 million edges), 100G (270 million of vertices, 1.7 billion edges) and 300G (800 million vertices, 5.3 billion edges), with 70% higher throughput and on average more than 6 times faster query performance. Galaxybase’s average response time, P50, P90, P95 and P99 response times all show better results than the previous record holder. In particular, the best average response time is greater than 41 times fasterand the best P90 response time is greater than 72 times faster.

No matter the requests of different data scaling factors under the same test tasks or on requests from different complexity for different test taskss, Galaxybase worked better. The larger the size of the datasets, the higher the differentiation, which fully demonstrates Galaxybase’s excellent ability to support large-scale data processing tasks.

Table 1: Summary of Galaxybase test results for different scale factors

Scale factor

Reference duration

Reference operations


meeting deadlines


2h 14m 35.740s





2h 05m 26.944s






2h 07m 24.645s




Notes: SF-30, SF-100, and SF-300 correspond to the original dataset size of 30G, 100G, and 300G respectively.

Galaxybase can support both online transaction processing (OLTP) and online analytical processing (OLAP), according to Yan ZHOU, CTO of CreateLink. Compared to other graph databases, Galaxybase features faster response time, higher throughput, and higher horizontal scalability, making it the perfect choice for a high-performance graph database on a large data set, as part of the growing demand for real-time analytics on massively connected enterprise data.

Galaxybase is written with Java and C++. It takes full advantage of runtime performance and memory control with C++, and coding efficiency and ease of troubleshooting with Java in developing complex and reliable systems. In terms of storage system design, Galaxybase uses an innovative proprietary native graph data store with custom optimization for index-free adjacency of graph data, allowing edge vertex queries to be completed extremely efficiently. The data store’s core engine does not rely on any third-party open source components, allowing the system to better optimize graph queries and graph computation in concert with the underlying storage layer. In the query execution layer, Galaxybase is able to efficiently organize data from memory through its proprietary memory allocation and management mechanism, while significantly reducing the JVM’s GC time by using the off-heap memory. Galaxybase provides a parallel iterative approach to traversing graphs, which uses multi-version checking to reduce lock contention. This approach adaptively allocates the number of threads for parallel iterations based on the number of neighbors during the neighboring iteration to achieve the best use of system resources.

Galaxybase provides a rich set of query interfaces and programming APIs, such as Java, Python, Golang, etc. It also supports the OpenCypher descriptive query language. For the audit test, Galaxybase performed the durability test using OpenCypher. For scenarios that require high system resource consumption and runtime performance, Galaxybase also provides PAR (Parameterized Algorithm Routine) API which allows users to implement custom procedures and functions running on the server side through Java code to better control the query execution process and seek extreme performance.

About Link

Founded in 2016, CreateLink Technology has become the leading provider of graph database software in China. Its Galaxybase product is a native distributed parallel graph platform that delivers superior performance on complex graph queries and algorithms. Earlier this year, with a team of experts from Sun Yat-sen University, Galaxybase has met the challenge of intelligent graph exploration of 5 trillion relationship graphs while maintaining security, performance, availability and data integrity, easily breaking the once difficult scale barrier with a relationship graph of over 1,000 trillion.

CreateLink actively promotes graphics technology in different industries with a diversity of application scenarios. So far, Galaxybase has served many high-profile customers in finance, energy and internet industries, with successful applications in anti-money laundering, fraud detection, power grid optimization, IT operations monitoring and maintenance, and other complex real-time decision-making scenarios, enabling its customers to unlock the value of big data by connecting the dots.

To check the LDBC-SNB test report and get more details about the report, please visit the GalaxyBase website.


Comments are closed.