Stay ahead in data processing and analytics with GBase 8a, a China-developed database.

By Chen Huaping

Previously, most systems within Fujian Rural Credit Union (Fujian RCU) ran on Oracle databases. As the business grew and data volumes increased, our data team referred to the case study of Agricultural Bank of China and gradually migrated our data warehouse from Oracle to GBase database, significantly improving batch processing performance – a case study also available on the GBase official website. During the pandemic last year, I was responsible for a system that planned to build a GBase 8a MPP Cluster database, migrating the data processing part from Oracle to GBase. To meet project requirements, I enrolled in the GBase 8a V86 training course and passed the certification exam. As the project progressed and I gained deeper understanding of GBase 8a, and to further strengthen my foundation and learn about the new features and improvements of V95 compared to V86, I signed up for the GBase 8a V95 certification course again this year and studied on my own after work.

Through these two courses, I felt the content was largely similar, with little difference. They basically covered three aspects: database technical architecture and principles, database installation, deployment and management, and database development and tuning. The courses were relatively foundational, but highly practical and hands-on. They should not be difficult for those with a background in other databases, especially MySQL. The data architecture of V86 and V95 is essentially the same, both adopting a Shared Nothing architecture. However, V95 adds a VC (Virtual Cluster) mechanism and supports multiple hash distribution columns, and has also enhanced security encryption. Since I haven't used V95, the other 40+ functional improvements remain to be explored.

In our project, we actually used GBase 8a MPP Cluster V86, and during usage we found that GBase delivers high batch processing performance for large data volumes. Previously, batch processing over 10 TB of data took eight to nine hours on Oracle, while GBase completed the job in less than two hours. Of course, during the database migration we also encountered various issues, such as physical machine installation and deployment, node scaling, significant differences from Oracle syntax, less-than-ideal enterprise manager tools, and low concurrent query performance. Some of these have been partially resolved or mitigated in actual use. Overall, GBase 8a database should be quite competitive in data analytics applications, and we should leverage its strength in distributed computing performance.

With the roll-out of the IT Application Innovation (Xinchuang) initiative, domestic databases will be further strengthened and adopted. As a key player in this initiative, GBase has successful deployment cases across various industries, and its stability, maintainability, and availability are continuously improving. I sincerely wish GBase database continued success and truly becomes a beacon of domestic databases.