Evolution of Taobao Backend Architecture: From Single‑Server to Cloud‑Native Microservices

The author, a senior architect, explains the evolution of Taobao's backend architecture from a single‑machine setup to a cloud‑native, micro‑service‑based system, focusing on how the architecture adapts to increasing concurrency from hundreds to millions of requests.

Basic Concepts : Distributed systems (multiple modules on different servers), high availability (fail‑over nodes), clusters (multiple servers acting as a single service), load balancing (even request distribution), and forward/reverse proxies (internal vs. external request routing).

Evolution Steps :

1st Evolution : Separate Tomcat and database onto different servers to eliminate resource contention.

2nd Evolution : Add local cache (e.g., Memcached) and distributed cache (Redis) to offload read traffic from the database.

3rd Evolution : Deploy multiple Tomcat instances behind a reverse‑proxy (Nginx/HAProxy) for load balancing.

4th Evolution : Implement database read/write separation using middleware such as Mycat.

5th Evolution : Split databases by business domain to reduce cross‑business contention.

6th Evolution : Partition large tables into smaller, hash‑routed tables.

7th Evolution : Use LVS or hardware F5 to load‑balance multiple Nginx layers.

8th Evolution : Apply DNS round‑robin to distribute traffic across data centers.

9th Evolution : Introduce NoSQL stores (HBase, Redis) and search engines (Elasticsearch) for analytics and full‑text search.

10th Evolution : Split the monolithic application into smaller, business‑oriented services.

11th Evolution : Extract common functionalities into independent micro‑services (using Dubbo, Spring Cloud, etc.).

12th Evolution : Adopt an Enterprise Service Bus (ESB) to unify service interfaces.

13th Evolution : Containerize services with Docker and orchestrate them with Kubernetes.

14th Evolution : Migrate the whole system to a public cloud (IaaS/PaaS/SaaS) for elastic resource provisioning.

Each stage resolves a specific bottleneck—CPU, memory, I/O, network, or operational complexity—while introducing new challenges such as data consistency, deployment overhead, and increased system complexity.

Design Principles summarized at the end include N+1 redundancy, rollback capability, feature toggles, monitoring, multi‑active data centers, using mature technologies, resource isolation, horizontal scalability, purchasing non‑core components, using commercial hardware, rapid iteration, and stateless service design.

The article concludes that architecture must evolve continuously to meet growing traffic and business demands, and that future topics like cross‑data‑center synchronization and distributed transactions deserve separate discussion.