GBase Data Warehouse Solution – selected in the 3rd Batch of Outstanding Financial IT Application Innovation Solutions

Recently, the Financial IT Application Innovation Ecosystem Laboratory officially released the "3rd Batch of Outstanding Financial IT Application Innovation Solutions". GBase, having been selected in the first and second batches, has once again been recognized, this time with four solutions, including one awarded as a "Laboratory Recommended Solution", reaffirming recognition and endorsement from authoritative institutions and industry clients. Let's take a look at the GBase database's "Data Warehouse Solution", featured in the 3rd Batch of Outstanding Financial IT Application Innovation Solutions~

Data Warehouse Solution

Aligned with a major state-owned bank's enterprise architecture deployment, this solution builds a "three horizontal, two vertical, one line" digital asset operation service system to solidify the bank's data governance. It leverages chip-based servers, an operating system, and a database to establish the enterprise data warehouse platform.

Application Scenarios

Data warehouse systems; customer marketing and expansion systems; financial data analysis systems; intra-city active-active; audit systems; data marts; ODS data application systems; storage management; regulatory analysis systems; remote active-active.

Technology Stack

GBase database + Hygon chips + NeoKylin operating system

Pain Points Addressed

Ø Breaking through single-core bottlenecks of domestically developed chips with multi-threaded parallelism

GBase database applies multi-threaded parallelism at every stage of SQL execution, harnessing multi-core capabilities of chips like Hygon to boost performance.

Ø Reducing performance loss from cross-NUMA Node memory access on NUMA architecture servers

A multi-instance deployment is adopted: two compute node instances are deployed on one Hygon chip server, each bound to an adjacent group of NUMA Nodes, mitigating cross-NUMA Node memory access overhead while improving server resource utilization. Two compute instances are deployed per chip server.

Ø Providing best practices for large-scale MPP clusters

Pioneering the large-scale use of domestically developed chip servers in MPP clusters, this solution provides replicable experience in cluster implementation, performance optimization, and O&M management. It has formed a comprehensive innovation implementation plan covering server configuration, deployment recommendations, and clear requirements for hardware such as disks and network cards.

Ø Multi-level high availability design for improved overall system availability

From the perspectives of data security and compute environment high availability, strategies are defined for node-innovation upgrade, data backup and recovery, and data reprocessing. During construction, GBase database's three-copy data storage and active-active cluster disaster recovery capabilities are utilized.

Innovation & Demonstration Highlights

GBase database is now used by over a hundred customers in the financial industry. One state-owned bank customer has deployed over 2,000 servers, processing nearly 20PB of total data, with historical data in the warehouse exceeding hundreds of terabytes and daily increments reaching tens of terabytes. Thousands of batch jobs have been launched, including nearly 2,000 core jobs that can be completed within an hour, delivering a cost-effective, highly reliable, and stable data warehouse platform.