Application Refactoring · System Migration
ICBC needed a faster, more elastic cloud architecture for digital banking workloads because its original serverless platform could not keep up with application growth, model release needs, and the bank's broader push toward cloud-native modernization.
Reported impact
ICBC improved elasticity, lowered infrastructure overhead, accelerated application and model delivery, and created a financial-grade cloud-native reference architecture for broader enterprise modernization.
ICBC rebuilt its financial-grade cloud platform around Baidu AI Cloud's Tianhe Stack and Function Compute to create a Serverless 2.0 architecture. The bank integrated the new function-computing core with its existing distributed systems and cloud platform, then used it to support BFF aggregation, server-side rendering, static assets, batch processing, database-update jobs, and rapid model publishing. Function Workflow lets teams compose larger batch tasks from individual functions, while Kafka-triggered execution and direct node connections support operational jobs that spike during peak release periods. By wrapping formerly heavyweight services in function-based interfaces, the bank avoided extra load-balancer deployments, improved elasticity, and gave project managers a faster path to put trained models such as investment backtesting into production. The platform is positioned as a supplement to microservices, helping ICBC modernize applications and lift resource efficiency without abandoning its broader distributed architecture.
Business requests from mini programs, H5, and other banking services are routed into function-based BFF, SSR, and static-resource components on the serverless platform. Batch and model-release jobs are triggered through workflows and Kafka events, and the resulting services or outputs are published back into ICBC's distributed banking environment for production use.
Application scenarios include BFF services, SSR, static resources, batch jobs, database updates, and model publishing (original: 主要包括BFF接口聚合服务、服务端渲染SSR、静态资源服务…批量任务…数据库安装场景…第三个场景是模型发布).
ICBC built a Serverless 2.0 architecture around Baidu's function-computing product on Tianhe Stack (original: 工商银行以百度智能云天合Stack的函数计算产品为核心打造了Serverless 函数计算平台技术架构2.0).
Migrating fund-combination backtesting models to function compute supports frequent and rapid releases (original: 通过将基金组合回测模型迁移到函数计算平台…需要频繁且快速的模型发布).
The platform integrates with ICBC's existing distributed and cloud-computing technology systems and provides a complete function engine, management, and delivery capability (original: 并与工商银行现有的分布式、云计算技术平台做对接，为应用提供完整的函数核心引擎、函数管理能力、开发交付能力).
The project improved model release speed by 10x and resource utilization by 80% (original: 模型发布速度提升10倍；资源利用率提升80%).
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Archived · Apr 28, 2026 CME Group CME Group the world’s leading derivatives marketplace, has built an enviable reputation for trust and reliability over its 175-year history. However, driven by a culture of innovation, it isn’t resting on its legacy and… 10+ hours of monthly productivity gains per developer Antigravity · Gemini CLI · Gemini Code Assist 
Alibaba Cloud Sample Archived · Jul 21, 2026 Go8 Technology Inc. As Go8 Technology expanded its portfolio of digital solutions, it required a scalable, reliable, and cost-efficient cloud platform to support growing customer demand and increasingly sophisticated workloads. The company… Reduced application latency from 100–250ms to 10–50ms through deployment on local Philippine infrastructure Application Load Balancer (ALB) · Container Service for Kubernetes (ACK) · Content Delivery Network (CDN) Google Cloud Sample Archived · Jun 11, 2026 InMobi InMobi needed to migrate a latency-sensitive advertising platform from another hyperscaler, preserve zero-downtime service at massive QPS, and create a more efficient real-time serving and analytics environment for both… 7 million peak qps Bigtable · Cloud Composer · Cloud SQL 
