Evolution of Banking Core System Architecture: Historical Review and Future Trends

System‑Budding Stage

In the early 1980s Chinese commercial banks began using micro‑computers for branch‑level savings processing. The "Micro‑computer Savings Business System" jointly developed by ICBC Chengdu Branch and Southwest Jiaotong University was deployed in January 1984. It supported query accounting, interest calculation, report generation and loss‑report handling, achieving the first electronic savings processing and markedly reducing staff workload.

The system used a single‑user architecture on VICTOR‑9000 hardware with DOS, file‑based storage, floppy‑disk backup and manual data‑consistency checks. Its logical flow consisted of data loading, printing, business processing, file copying and interest calculation, with a main program that invoked function modules based on user input.

Centralized Architecture Stage

Mid‑late 1980s marked the introduction of the SAFEII system, the first large‑scale mainframe‑based core system. Deployed by ICBC in 1987‑1988, SAFEII processed retail and corporate business, supported same‑city inter‑bank transfers and enabled regional networking via a "big‑machine extension" project.

SAFEII employed a client‑server structure: thin clients on independent workstations and a robust mainframe server with CICS middleware for layered development and transaction‑based data consistency. The logical structure was hierarchical, modular and parameterized, with control tables guiding teller verification and transaction routing.

Limitations of SAFEII—insufficient support for diversified business, inability to interconnect systems, performance bottlenecks under high load, and high maintenance cost of assembly‑language code—prompted the next generation. In 1999 ICBC launched the "9991 Project" and the CB2000 core system, introducing front‑back separation, relational databases, message middleware and a two‑site‑three‑center disaster‑recovery model. Smaller banks meanwhile adopted small‑computer platforms, forming a distinct technical stack.

Distributed Architecture Stage

Around 2010 banks began building distributed core systems on small‑computer clusters. Examples include Bohai Bank’s IBM small‑computer cluster (2006), China Merchants Bank’s AS400 cluster (2013) and Postal Savings Bank’s open‑system cluster (2013). Cluster architectures typically consist of access, application and data layers. They improve horizontal scalability and concurrency via load balancing, fault isolation and traffic control. Data‑layer implementations vary: Bohai Bank used DB2 pureScale; China Merchants Bank introduced vertical and horizontal sharding with custom transaction‑penetration and dual‑machine protection mechanisms; Postal Savings Bank employed data‑sharding and transaction‑compensation mechanisms to ensure fund safety.

Although cluster‑based cores enhanced performance, they still faced high concurrency limits, scalability constraints and single‑point‑failure risks.

From circa 2015 the "distributed unit‑based" architecture emerged, decomposing the system into self‑contained units (micro‑services) that each handle specific business functions and data partitions. Banks such as Bohai (2021 "Blue Sea Project") and Postal Savings (2023) have deployed this architecture.

Technical characteristics include reliance on PC servers and open‑source components, addition of HTTP calls alongside local and RPC calls, fine‑grained scaling at service, node and unit levels, and fault‑containment mechanisms that limit impact to affected units. Logical designs remain hierarchical, modular and parameterized, with micro‑service orchestration enabling flexible composition and rapid iteration.

Challenges of unit‑based designs involve complex data‑layer sharding, distributed transaction handling (SAGA, TCC), increased operational entities and the need for specialized modules such as transaction routing.

Future Evolution Trends

Accelerating domestic‑technology adoption for full‑stack autonomy – National policies (14th Five‑Year Plan) mandate financial‑industry information security and promote indigenous hardware/software. By end‑2025 Postal Savings Bank achieved full‑stack autonomous control of core services [12] .

Applying cloud‑native technologies to improve efficiency and stability – DevOps shortens delivery cycles; containerization enables auto‑scaling; service mesh, serverless and observability enhance resilience. Construction Bank’s distributed core system exemplifies cloud‑native deployment [13] .

Introducing intelligent operations (AIOps) to cut maintenance costs – Real‑time data collection, predictive analytics and automated remediation raise fault detection by 50 % (Agricultural Bank case) [14] .

Adopting real‑time data warehouses for enhanced data services – To meet low‑latency analytics, banks integrate Flink‑based real‑time computation pipelines, as demonstrated by China Bank [15] .

Deepening micro‑service architecture for maintainability – “Thin core” concepts evolve into decoupled micro‑service layers, enabling agile, flexible, cost‑effective system evolution; ICBC’s ECOS project illustrates a shift toward an open‑ecosystem, “core‑less” model [16] .

Integrating large‑model AI to reshape user interaction – AI is applied at the interface layer to provide multimodal (voice, image) interaction, intent recognition and personalized product recommendation.

Code example

[1] 吴峰,陈展荣.中国工商银行广东省分行科技发展简史(1984～2002年)[J].中国金融电脑, 2019(2):6.DOI:CNKI:SUN:ZGJN.0.2019-02-019.
[2] 邓幼强,龙志和.微型计算机储蓄业务处理系统[J].管理现代化, 1984(3):4.DOI:CNKI:SUN:GLXX.0.1984-03-006.
[3] 苏文力.研究评估SAFEII,建立新型金融结构[J].中国金融电脑, 1995(3):22-24.DOI:CNKI:SUN:ZGJN.0.1995-03-005.
[4] 罗毅.数据集中建伟业智慧银行谱新章[J].中国金融电脑, 2019(9):4.DOI:CNKI:SUN:ZGJN.0.2019-09-005.
[5] 邬德云,杨文.中国金融IT的发展研究[J].上海金融学院学报, 2010(6):9.DOI:10.3969/j.issn.1673-680X.2010.06.004.
[6] 王飞鹏,李宁宁,游晋澜,等.渤观约取 海润万物——新核心系统分布式单元化架构的应用实践[J].金融电子化, 2022(11):88-90.
[7] 杨勉.招商银行新一代分布式核心系统建设[J].中国金融电脑, 2016, 12(12):40-40.DOI:CNKI:SUN:ZGJN.0.2016-12-007.
[8] 李朝晖.邮储特色的分布式系统架构模式[J].金融电子化, 2017(5):3.DOI:CNKI:SUN:JRDZ.0.2017-05-016.
[9] 商业银行分布式业务核心系统架构研究[J].金融电子化, 2020(7):54-56.
[10] 佚名.邮储银行新一代个人业务核心系统建设圆满收官[J].中国金融电脑, 2023(2):95-95.
[11] 中国农业银行.农业银行顺利实施大型主机关停下电[EB/OL].2025-03-30.https://mp.weixin.qq.com/s/vI-GAJo51stV2opFqlcgUQ.
[12] 朱礼华.以自主可控筑牢金融"五篇大文章"根基——邮储银行核心系统基础设施全栈适配自主可控标准实践[J].中国金融电脑, 2025(12):40-43.
[13] 金磐石.打造云原生能力赋能新金融行动[J].中国金融电脑, 2022(5):7-10.
[14] 李树尉.数智赋能，运维转型——农业银行智能运维建设和应用实践[J].金融电子化, 2024(002):000.
[15] 宋首文.实时计算在商业银行数字化转型中的应用及建议[J].中国金融电脑, 2023(5):35-40.DOI:10.3969/j.issn.1001-0734.2023.05.007.
[16] 杨一军.工商银行去核心化开放生态银行建设[J].中国金融电脑, 2022(2):43-46.