Web Cluster Architecture

Web Cluster Architecture

Overall architecture:

Detailed architecture:

1. Client Layer

• Browsers, mobile apps, and third-party systems access the platform through HTTP/HTTPS.

2. Load Balancing Layer

• Nginx cluster:
  • Reverse proxy and backend load balancing.
  • SSL termination and static resource caching.
  • Horizontal scaling for high availability.

3. Application Service Layer

• Spring Boot service cluster:
  • Provides RESTful APIs and handles business logic.
  • Multiple instances are deployed horizontally and balanced through Nginx.

4. Message Queue Layer

• Kafka cluster:
  • Topics can be split into multiple partitions, such as Partition 1 and Partition 2.
  • Consumes device data from the FMQ cluster, including MQTT/TCP/UDP service clusters.
  • Device gateway services write data into Kafka.
  • Spring Boot services consume Kafka data and process device commands, alarms, and other business events.

5. Data Processing Layer

• Rule engine:
  • LiteFlow executes scene-automation rules such as device linkage and alarm triggers.
  • Ignite stores and manages LiteFlow rule configuration and supports dynamic updates.
• Distributed scheduling:
  • Quartz cluster runs Modbus polling jobs and coordinates cluster tasks through database locks.

6. Cache and Metadata Layer

• Distributed cache:
  • Ignite cluster caches device metadata such as device IDs, status, and configuration.
  • Integration with Spring Boot provides low-latency metadata access.
• Redis primary-replica architecture:
  • The primary node handles writes such as session state and real-time counters.
  • Replica nodes handle reads and can be combined with Sentinel for automatic failover.

7. Data Storage Layer

• Relational database:
  • MySQL primary-replica architecture, where the primary handles transactional writes and replicas handle queries.
• Time-series database:
  • TDengine stores historical device data such as temperature and humidity and supports efficient time-series queries.

8. Device Access Layer

• Protocol gateway:
  • Supports MQTT/TCP/UDP protocol access.
  • Forwards device data to the Kafka cluster.

Key Design Points

1. High availability: critical components such as Nginx, Kafka, MySQL, and Redis use cluster or primary-replica deployment.
2. Data flow: device data -> MQTT/TCP/UDP gateway -> Kafka -> Spring Boot services -> TDengine/MySQL.
3. Performance: Ignite caches device metadata to reduce MySQL pressure, and TDengine optimizes time-series storage and queries.
4. Scalability: Kafka partitions and Nginx load balancing provide horizontal scaling.