Transforming BFF Development: Ant Group’s Needle Node FaaS Platform Explained

Why: Ant BFF Challenges and Thoughts

1.1 BFF Development Challenges and Bottlenecks

In BFF development we encounter issues such as large BFF applications being blocked by branch iteration, high collaboration costs for module compatibility, and increased quality and release risk.

Previous efforts on code organization (enterprise framework), workflow (gitflow) and collaboration tools (scrum) mitigated some problems, but application‑level delivery granularity still causes code coupling and unnecessary collaboration overhead.

BFFs are essentially function‑level interfaces; moving to a function‑level delivery platform enables fast write, test, and deploy, reducing framework dependence and collaboration friction.

However, Ant’s financial‑grade, high‑traffic scenarios require a safe and reliable FaaS solution, which many competitors lack due to migration cost and immature community solutions.

1.2 Progressive Approach to Overcome FaaS Adoption Barriers

By decomposing FaaS characteristics we leverage rapid development and dynamic deployment at function granularity, and serverless elasticity to eliminate operations.

We embed an isolation‑aware, dynamically deployable container solution as a plugin to the existing application framework, reusing hardware resources while providing function‑level deployment.

The solution is transitional; we also plan gradual migration to a standard FaaS infrastructure.

How: Node FaaS Platform Key Design and Considerations

2.0 Enterprise‑Level Platform Design Methodology

The journey from idea to enterprise platform involves three steps: technology selection, architectural design, and iterative optimization.

Technology selection starts with identifying the decisive technology—function runtime containers—and balancing reliability, efficiency, and optional performance/isolation requirements.

Architecture design separates a control plane (developer operations) and a runtime plane (service execution), using a layered approach to minimize coupling and dependencies.

Four lifecycle stages—development, quality, deployment, and operation—are supported by tools such as an online document‑style editor for remote BFF development and debugging.

2.1 Runtime Container Technology Selection

We compare VM and other options; VM offers advantages in data isolation, startup performance, and development cost, making it the preferred choice despite slower execution due to lack of TurboFan JIT optimizations.

2.2 Architecture Design Philosophy

Needle follows a PaaS‑style architecture, with a control plane for developer tooling and a runtime plane for reliable service delivery, emphasizing low coupling and automated file‑based triggers.

2.3 High‑Availability Reinforcement

High availability is mandatory for national‑level financial platforms. Needle adopts “gray‑scale, monitorable, emergency‑ready” principles, implementing granular rollout, real‑time health dashboards, and rapid rollback or incremental updates within minutes.

2.4 Phase Outcomes

Development efficiency improved from day‑level to minute‑level delivery; over six months the platform saved 400+ person‑days, serving 40+ business applications across Ant Wealth and Insurance.

Next: Future Technical Exploration

Future work focuses on reducing request latency, lowering per‑QPS cost, and achieving full FaaS compatibility with backward‑compatible migration paths.