Evolution of E‑commerce System Architecture: From Monolithic to Unitized Design

In the early 1.0 era, the e‑commerce site was a single web project containing all modules and a single database with all tables; as traffic grew, the bottleneck became the application server connection limit (e.g., Tomcat threads), prompting horizontal scaling of web servers.

During the 2.0 era, adding more web servers shifted the bottleneck to the database, limited by maximum connections (e.g., 8000) and large table operation latency; solutions included read/write separation with master‑slave replication and caching, or vertical business splitting when read/write ratios were extreme.

The 3.0 era introduced read/write separation while still maintaining a vertical architecture; however, the need for faster iteration led to vertical module extraction, creating independent services for large modules.

In the 4.0 era, business vertical splitting was applied, and CDN caching was added to improve user experience by serving content from the nearest edge node.

Further scaling in 4.1 introduced sharding and database partitioning to handle tables with tens of millions to billions of rows, alleviating read/write pressure and connection limits.

Version 4.2 added a dual‑datacenter (same‑city) deployment with dedicated links, reducing latency to a few milliseconds but still facing cross‑datacenter database access overhead.

Version 4.3 highlighted the remaining challenges of database connection limits and the need for a unit‑based (SET) architecture, where core systems are divided into independent units that handle traffic based on routing rules (e.g., user ID), achieving high cohesion and self‑containment.

The 5.0 unitized architecture routes traffic from the access layer to different units, each containing core services and matching data‑sharding logic, solving database connection bottlenecks while still requiring some cross‑unit calls for non‑splittable business functions.

Reference articles include Li Zhihui’s "Large‑Scale Website Architecture" and several online blogs.

Additional promotional content (ChatGPT community, interview material, and related offers) is omitted from this technical summary.