Baidu Contract Advertising Platform: Microservices Architecture Evolution and Best Practices

This article explores the technical challenges and architectural evolution of B-end (B2B) business systems, using Baidu's Contract Advertising Platform as a case study. The author examines how the platform transitioned from a monolithic architecture to microservices, addressing complex business scenarios involving hundreds of advertising and delivery types.

Contract Advertising Concept: Unlike real-time bidding advertising, contract advertising pre-negotiates advertising prices. The delivery process consists of four steps: inquiry → order placement → delivery → settlement. This model is more suitable for brand display advertising and involves heavier pre-delivery业务流程 compared to bidding advertising.

Business Development Phases: The platform went through three stages: 1) Rapid growth period (2011-2018) with independent delivery platforms for each product line; 2) Integration phase (2019+) seeking product consolidation and process unification through the Tianqi Platform; 3) Complex marketing scenario support (2021+) enabling integrated sales of multiple advertising products under one contract.

Architecture Evolution: The 1.0 monolithic architecture was a "chimney-style" structure where each product line operated independently, leading to information silos. The 2.0 microservices architecture adopted Domain-Driven Design (DDD) principles, dividing the system into four layers: Business Frontend, Business Middle Platform, Technical Components (PaaS), and Infrastructure (IaaS).

Technical Implementation:

Module Structure: Using COLA architecture with DDD, CQRS, and dependency inversion principles. Modules are organized by domain first, then by function.

Service Governance: Focuses on three aspects: availability (performance optimization, self-healing), observability (business entity lifecycle tracking), and anti-corruption (preventing architecture decay).

Service Iteration: Emphasizes forward compatibility, automation testing, and service mocking capabilities to achieve service autonomy and isolation.

Key Insights: The biggest technical challenge in B-end systems is managing business complexity. A good business architecture should improve delivery efficiency, operations efficiency, and evolution efficiency. The author concludes that no architecture is perfect—only one that evolves with actual business needs.