Evolution of Large‑Scale Website Architecture: From PHP to Java, Distributed Systems, Mobile, Mid‑Platform and Internationalization

When a startup begins, it often runs on just two machines—a web server and a database server—hosting all functional modules in a single application; as traffic and business grow, this monolithic approach quickly becomes insufficient.

The evolution of the website architecture is described in five main stages: migration from PHP to Java, distributed transformation, wireless (mobile) transformation, mid‑platform transformation, and internationalization.

Stage 1 – From PHP to Java Early sites commonly used the LAMP stack (Linux, Apache, MySQL, PHP), which allowed rapid development but suffered from poor scalability and limited support for complex enterprise logic. Companies switch to Java for its stronger type system, richer ecosystem, and better suitability for large‑scale business applications. Factors influencing this decision include language characteristics, availability of skilled developers, and maturity of tooling and middleware. After the migration, the stack became WebX + EJB + iBatis + JBoss + Oracle (later Spring), leading to storage bottlenecks that were addressed with IBM mainframes, Oracle databases, EMC high‑end storage, database sharding, distributed cache (Tair), distributed file system (TFS), and CDN.

Stage 2 – Distributed Transformation Distributed transformation aims to make the system stateless or contain state within bounded contexts to enable horizontal scaling. This involves micro‑service decomposition, establishing a distributed RPC framework, asynchronous messaging, distributed data layers, distributed file systems, and service discovery/registration. Session handling must also be addressed. The business was split into commodity services such as product center, transaction center, user center, and shop center, forming a service‑oriented architecture (see diagram).

Stage 3 – Wireless (Mobile) Transformation By 2013, mobile technology had become dominant. Previously, mobile sites were copies of PC sites, leading to duplicated effort. The "All‑in Wireless" project unified PC and mobile front‑ends, modularized and componentized the system, and experimented with Node.js to improve front‑end development efficiency. Although Node generated enthusiasm, its long‑term impact was limited, and further details are discussed in later chapters.

Stage 4 – Mid‑Platform Transformation The mid‑platform concept, inspired by the U.S. military’s command structure, addresses the inefficiencies of a sprawling, fragmented system after years of growth. By extracting common capabilities into a shared platform, organizations improve development efficiency, reduce duplication, and lower innovation costs.

Stage 5 – Internationalization Two approaches exist: deploying independent codebases for each region, or true internationalization that shares a single codebase across multiple locales. Internationalization must solve multilingual support, time‑zone handling, data routing, and global data synchronization and replication.

Excerpted from the preface of Large‑Scale Website Architecture Evolution and Performance Optimization by Xu Lingbo, published by Electronic Industry Press, June.

Source: http://www.broadview.com.cn/article/419460

Architecture Digest ID: ArchDigest Internet Application Architecture | Architecture Technology | Large‑Scale Websites | Big Data | Machine Learning