Industrial and Commercial Bank of China's IT Architecture Transformation: A MySQL Distributed Enterprise Solution

Background and Challenges – The bank’s legacy IT stack relied on mainframes and DB2, which could not meet the growing demands for processing capacity, 24/7 availability, rapid delivery, and cost control in the era of internet finance.

Core Demands and Strategy – The transformation aimed to achieve high concurrency, scalable storage, two‑site three‑center disaster recovery, reduced licensing costs, and automated, open‑source‑based operations.

Roadmap – The three‑year plan consisted of a prototype phase (2016‑2017), a research and pilot phase (2017), and large‑scale implementation and promotion (starting 2018).

Technology Selection – After extensive evaluation, the bank chose an open‑source MySQL‑based distributed OLTP solution, rejecting NewSQL due to maturity concerns and focusing on a distributed architecture to address scalability and high‑availability requirements.

Distributed Technology Stack – The stack includes distributed services, transaction frameworks, batch processing, caching, data reconciliation, messaging, configuration management, and development/operations tools.

MySQL High‑Availability Design – Leveraging semi‑synchronous replication, multi‑master setups, and automatic failover, the solution achieves RPO = 0 and sub‑minute RTO across two Shanghai sites and a Beijing disaster‑recovery center.

Implementation Phases – Phase 1 built the prototype; Phase 2 conducted research and pilot deployments; Phase 3 rolled out the solution enterprise‑wide, integrating the DBLE distributed middleware for sharding and cross‑database queries.

Operations and Automation – A unified management platform provides automated installation, monitoring, alerting, fault analysis, health scoring, and both manual and automatic failover capabilities.

Optimization and Improvements – Continuous enhancements include expanding from 1‑primary‑2‑standby to 1‑primary‑3‑standby configurations, adopting SSD storage to eliminate I/O bottlenecks, and containerizing MySQL to improve resource utilization by 4‑5×.

Results – Over 120 applications, 2,000+ servers, and 2,500+ MySQL instances were migrated, supporting daily transaction volumes of up to 700 million and peak TPS of tens of thousands, while reducing hardware and licensing costs.

Future Work – The bank plans to advance cloud‑native services, real‑time data replication, Oracle migration, and further exploration of distributed database technologies to ensure autonomous, resilient capabilities.