How Infrastructure Works

The architecture of the Tingyun Infrastructure is shown in the diagram below:

Infra agents are divided into host agents and component agents. Host agents are responsible for monitoring host CPU, memory, disk, and network metrics, process metrics, and Docker container metrics. Component agents mainly connect to related components remotely to obtain performance metrics of related components.

After the agent starts, it registers itself to the Collector. The Collector pulls data from the agent and uploads it to DC. After DC performs data authentication, it sends the data to Kafka. Druid pulls data from Kafka for storage.

There is a heartbeat between Collector and DC. The Collector is stateless and can be horizontally scaled for high availability and load balancing.