How Tencent Cloud’s Unitized Architecture Boosts Microservice Scalability and High Availability

Unitized Architecture Overview

In traditional layered micro‑service architectures each layer uses its own partitioning algorithm, causing unpredictable cross‑zone or cross‑region calls. Unitized architecture eliminates this uncertainty by pre‑defining call paths and binding compute resources and data resources with a common logical tag. The result is a set of standardized processing units that encapsulate both compute and a single data store.

Unit Characteristics and Types

Each unit contains a group of compute instances and exactly one data resource, both labeled with the same tag.

A system typically defines 4‑12 units, but the number can be larger depending on business scale.

Units are classified by the business role they host:

Access/Gateway Unit (ADU) – traffic ingress and egress.

Standard Processing Unit (SDU) – core business logic.

Local Unit (LDU) – single‑AZ shared services.

Regional Unit (RDU) – same‑city shared services.

Global Unit (GDU) – cross‑region services in active‑active scenarios.

Why Adopt Unitization?

When a system grows, the following problems become acute:

Database connection explosion : Each unit binds to a dedicated database, allowing connection limits to be managed per unit and enabling distributed scaling.

Complex scaling operations : Scaling is performed at the unit level, making capacity planning predictable.

Cross‑datacenter latency and instability : Unit traffic forms a closed loop, minimizing cross‑center calls.

Active‑active multi‑region deployment : Unitization provides a practical path to multi‑region high availability.

Tencent Cloud Four‑Layer Unitized Model

The model consists of Access, Application, Data, and Infrastructure layers. The Access Layer receives requests, identifies the target unit tag, and forwards traffic to the corresponding Application Unit. Application Units are split by customer, achieving per‑customer transaction isolation.

Tencent Service Framework (TSF) Core Capabilities

Standardization : Unified service entry, registration, configuration, and deployment experience.

Technology diversity : Supports Spring Cloud, Dubbo, Service Mesh, gRPC, and other mainstream frameworks.

Full‑stack governance : Permission control, performance tuning, and end‑to‑end service management.

Key Design Points

Unit Routing

Clients attach a logical tag to each request (e.g., customer ID). The unitized gateway computes the target unit, determines whether the call is intra‑unit, intra‑center, or inter‑center, and routes accordingly. The routing order is:

Intra‑unit call (same unit).

Intra‑center call (same datacenter, different unit).

Inter‑center call (different datacenter).

Unit Gray Release

Gray release is realized by defining one or two dedicated gray units and a gray‑dimension table (e.g., customer ID, tag). Before routing, the gateway checks the gray table; matching requests are sent to the gray unit, enabling isolated rollout of new versions or configurations.

Unit Disaster Recovery

For active‑active scenarios each unit has a backup unit recorded in a mutual‑backup table. When a primary unit fails, traffic is switched to its backup and the associated database replica is promoted to primary. After the failed unit recovers, traffic can be switched back.

Implementation Workflow

Plan the number of units and provision unitized products: unitized gateway, microservice platform, message queue, and database.

Define unit rules (tag mapping, backup relationships, gray‑unit definitions) and push them to the respective components.

Configure optional gray units and gray‑release rules.

Enable TSF disaster‑recovery settings: one‑click failover, simulation drills, and unified monitoring/alerting for cross‑unit or cross‑datacenter requests.

Practical Example of Unit Routing Logic

// Pseudocode executed by the unitized gateway
function routeRequest(request) {
    tag = request.headers["X-Unit-Tag"];
    targetUnit = unitMap[tag]; // unitMap is populated from unit rules
    if (isInSameUnit(request, targetUnit)) {
        forward(request, targetUnit);
    } else if (isInSameCenter(request, targetUnit)) {
        forwardCrossUnit(request, targetUnit);
    } else {
        forwardCrossCenter(request, targetUnit);
    }
}

Core TSF Unitized Product Matrix

TSF provides three major capability groups for unitized deployments:

Unit Management : unit rule configuration, tag‑based routing, gray‑release, backup‑unit tables.

High‑Availability Disaster Recovery : one‑click failover, backup‑unit promotion, disaster‑simulation drills.

Efficient Operations : unified monitoring, cross‑unit/center alerting, standardized scaling procedures.

Key Architectural Diagrams