Microservices, Service Mesh, and Middle Platform: Architectural Evolution and Practices

On May 25, an Internet Architecture Technology Salon concluded with a presentation by Tencent senior engineer Han Xin, who shared insights on technology architecture, practical cases, serverless cloud functions, and massive storage system architectures.

The talk covered three main directions: the evolution of cloud computing architecture, the concept of a middle platform, and practical experiences with microservice middle platforms.

Microservices, originally proposed by Martin Fowler, are described as a software architecture that decomposes an application into small, autonomous services organized by business domains. This approach reduces development difficulty, enhances scalability, and supports agile development, continuous integration, and delivery. The speaker contrasted monolithic applications with microservices, noting the shift from tightly coupled modules to loosely coupled, single‑purpose services, often organized in “two‑pizza” teams.

The evolution of architecture was traced from early monolithic J2EE layers (presentation, business, data access) through SEB/SOAP, to modern containers, cloud‑native, agile, and DevOps practices. Drivers for architectural change include rapid internet growth, the need to serve massive user bases (e.g., insurance moving to online channels), agile delivery demands, and cloud‑based infrastructure that lowers deployment and management costs.

Cloud computing was broken down into three layers: IaaS (infrastructure), PaaS (platform), and SaaS (software). The speaker emphasized that as one moves upward, business relevance increases while standardization decreases.

Containers are highlighted as a key enabler for microservices, providing isolated runtime environments that bridge the gap between IaaS and PaaS. Service governance topics such as RPC frameworks, REST APIs, service authentication, routing, rate limiting, capacity planning, circuit breaking, and degradation were discussed.

Service Mesh was introduced as an emerging technology built on top of containers to provide advanced traffic management and observability, though current community implementations have limitations in protocol support and may introduce bottlenecks.

The concept of a middle platform (中台) was explained as a layer between front‑end and back‑end, aiming to provide reusable capabilities, reduce system cost, and improve organizational efficiency. Three tiers of a middle platform were described: cloud infrastructure layer, technical middle platform (providing services such as rate limiting, routing, distributed caching), and business middle platform (industry‑specific solutions).

Technical middle platform challenges include integrating frameworks, runtimes, and support services, essentially “simplifying complexity” and achieving deep abstraction.

TSF (Tencent Service Framework) was presented as a microservice‑based technical middle platform offering core capabilities such as microservice frameworks, DevOps, service governance, data‑driven operations, and service management. The architecture includes external users, third‑party services, CLB, API gateway, and supports both virtual machines and containers.

Service governance mechanisms were detailed: authentication, routing (weight and tag based), rate limiting, capacity tiers, circuit breaking, isolation, and degradation (API‑level and tag‑level). Resource management topics covered containers, namespaces, deployment groups, application packaging, version control, distributed configuration, gray releases, multi‑environment deployment, global logging, and API governance via annotations.

Additional discussions included microservice gateways, distributed transactions (message‑queue based, TCC, and framework‑managed transactions), and operational capabilities such as gray releases, full‑link pressure testing, dependency analysis, slow‑query analysis, serverless support, and large‑screen operation monitoring.

The speaker concluded that a good technical architecture should be simple, fast, and efficient, serving business needs directly. The talk emphasized that technology is a tool to generate business value, and the best architecture aligns closely with business requirements.