Baidu Code Development Collaboration Platform: Architecture, Evolution, and Enterprise SaaS Practices

Baidu introduced an end‑to‑end R&D toolchain—including a project management platform, code development collaboration platform, and continuous delivery platform—to boost overall development efficiency across demand, development, and delivery stages.

The code management goal focuses on the development phase, and Baidu proposes a five‑level pyramid model covering code hosting, collaborative development, DevOps support, data‑driven process improvement, and engineering culture.

Key challenges stem from a workforce of tens of thousands, nearly 100,000 projects, massive weekly code changes, and the need for automated quality checks, distributed compilation, and continuous integration, leading to three core issues: code quality, large‑scale collaboration, and service stability.

The platform addresses five problem areas: code hosting, collaborative development, code quality, code security & openness, and R&D improvement, providing features such as automated reviews, security scanning, workflow support (branch, trunk, feature, Git‑flow), and data‑driven process metrics.

Architecture evolution is described in four stages:

Product inception: a simple master‑slave single‑instance design with RAID, DRBD replication, and KeepAlived for failover.

Product growth: adoption of large memory, SSDs, I/O, network, and cache optimizations; evaluation of distributed storage vs. data sharding, ultimately choosing sharding for high I/O workloads.

Product maturity: read‑write separation, proxy‑based routing, 10 GbE bandwidth, additional read‑only nodes, and a custom high‑availability scheme based on proxy heartbeats and voting.

Enterprise SaaS: a dedicated cloud solution with multi‑tenant isolation at the access, application, and resource layers, supporting Docker and VM deployments.

Overall, the platform follows a micro‑service architecture built on Baidu's internal framework, comprising:

Access services (HTTP/SSH proxies, API gateway) with high‑availability load balancing.

Unified security and authentication.

Service center for registration, discovery, routing, configuration, and resilience.

Open services (Webhooks, Plugins) for CI integration and automated code checks.

Business services offering code management, browsing, review, search, and scanning.

Git cluster with triple backup, data sharding, read‑write separation, and automatic failover.

Foundational services such as databases, indexing, caching, user management, notifications, and storage.

For enterprise‑grade SaaS, Baidu provides a dedicated cloud that isolates tenants via unified proxy authentication, shared service layers with access‑control isolation, and physical resource separation using Docker or virtual machines, supporting both mixed and independent deployments.

The platform successfully improves capacity, performance, and reliability, yet challenges remain: data‑sharding complexity, operational overhead, and limitations under extreme write‑heavy workloads, highlighting the need for continuously evolving architectural solutions.