GBase 8a V9 Passes the Authoritative TPC-DS Benchmark

On December 23, the TPC (Transaction Processing Performance Council), the world's most recognized non-profit benchmark standardization organization for data management systems, officially published the TPC-DS 10000 Scale test results for the GBase 8a V9 MPP database. Using only 8 servers, GBase 8a V9 achieved an outstanding performance score of 8,944,478 QphDS, using the fewest test nodes among all tested vendors, with the highest per-core performance and the best data loading performance.

As enterprises undergo digital transformation and data grows explosively, increasing demands for data analysis and data mining have become critical benchmarks for evaluating big data solutions. The TPC-DS benchmark officially introduced by the TPC includes comprehensive assessments of statistics, report generation, online queries, and data mining over large data sets, using data consistent with real-world scenarios. It closely mirrors actual operations and is currently the core benchmark globally for measuring the maturity and competitiveness of a data warehouse. In this test, the GBase 8a V9, a domestically developed MPP database product, demonstrated exceptional capabilities.

Fewest Test Nodes:

Analytical MPP clusters are typically deployed on physical servers. For this test, GBase used physical servers as the foundational resource. Among all vendors that completed the benchmark, it used the fewest physical nodes—only 8—to finish all testing, and ranked first in performance on physical servers among all vendors. The GBase 8a V9 cluster, as an outstanding Chinese analytical database product, fully leverages the performance of physical machines, delivering optimal results in on-premises physical deployments. As a Shared-Nothing distributed database cluster system, GBase 8a V9's performance scales linearly with the number of nodes. It also supports online cluster expansion and contraction, allowing users to add or remove nodes without service interruption, meeting on-demand elastic cluster deployment requirements.

Highest Per-Core Performance:

With the continuous evolution of CPU technology, modern CPUs adopt multiple NUMA groups and ever-increasing core counts, posing significant challenges to database software architecture and concurrency. GBase 8a V9 employs a multi-instance architecture, binding instances to NUMA nodes for precise resource allocation, reducing cross-NUMA memory access overhead, and enhancing internal concurrency within each instance to effectively utilize CPU resources. This benchmark score was achieved on Intel(R) Xeon(R) Gold 6230R CPU @ 2.10GHz processors, with an average per-core performance of up to 1,118,059, fully exploiting the powerful single-core computing capabilities of Intel Xeon(R) CPUs.

Best Loading Performance:

GBase 8a V9 uses multi-point transmission technology, with every node in the cluster participating in data parsing and loading, enabling loading performance to scale linearly as nodes increase. In this test, only 8 nodes achieved a loading speed of 33TB per hour. In massive data scenarios, it efficiently meets users' needs for fast data ingestion at any scale.

GBase 8a V9 supports parallel loading from multiple data sources (FTP, HTTP, HDFS, local files, etc.), as well as flat files and gzip, lzo, snappy compressed formats. A single node loads over 300GB/hour per table, and the cluster loading performance exceeds 30TB/hour.

GBase 8a V9, as a mature analytical MPP database independently developed by GBase for the big data era, not only marks GBase as one of the Chinese database vendors passing the TPC-DS benchmark, but also represents the product with the fewest physical nodes, fewest total physical CPU cores, and least total memory among all tested big data solutions, while achieving the highest loading performance.