Overview

The cluster management component mainly serves cluster scenarios and is divided into the following two subcomponents based on application scenarios:

Table 1 Component types

Component Name

Application Scenario

Capability Description

Controller

O&M management during cluster running

The Controller completes the service status management and control of all Servers in the cluster, PD identity management and decision-making, and resource management and decision-making. It is the status controller and decision-making center of the entire cluster. For details, see Controller.

Coordinator

Inference request entry during cluster running

The Coordinator is the entry of user inference requests. It receives high-concurrency inference requests, schedules, manages, and forwards requests. It is the data request entry of the entire cluster. For details, see Coordinator.

Supported Features

Feature

Atlas 300I Duo inference card+Atlas 800 inference server (model 3000)

Atlas 800I A2 inference server

Atlas 800I A3 SuperPoD Server

Service deployment on a single node

Supported

Supported

Supported

Single-node deployment

Not supported

Supported

Supported

Multi-node deployment

Not supported

Supported

Not supported