Skip to main content
Version: v2512

Advanced Setup

This page explains complex configuration methods for AIBooster in on-premise environments.

Choosing a Configuration Pattern​

ConfigurationFeaturesRecommended Use
Single NodeComplete on one machineValidation, learning, small PoC
Multi NodeDistributed across multiple machinesProduction operation, large-scale clusters

Single Node Configuration​

A configuration where both AIBooster Server and Agent run on a single machine.

Pattern 1: Minimal Configuration for Validation​

single-node-1

Install both AIBooster Server and AIBooster Agent on a single workstation or server equipped with GPU. Connect a monitor and open the dashboard locally to check performance information. This is the shortest route when you "just want to get it running" on offline validation machines or benchmark machines. No network configuration is required.

Pattern 2: Multi-User Configuration​

single-node-2

Install both AIBooster Server and AIBooster Agent on a single workstation or server equipped with GPU. Users view the dashboard provided by the server from their personal PCs through a browser via TCP port 3000. Ideal for small PoCs where multiple people want to view the dashboard.

Multi Node Configuration​

A configuration for production operation where AIBooster Server and Agents are distributed across multiple machines.

multi-node-1

Install AIBooster Server on the management node, and install AIBooster Agent on each GPU compute node. Users view the dashboard provided by the management node from their personal PCs through a browser via TCP port 3000. This is the recommended configuration for many GPU cluster server systems.

Pattern 2: Co-located with Compute Node​

multi-node-2

If there is no dedicated management node, select one GPU-equipped node and install both AIBooster Server and its dedicated AIBooster Agent there. Install only the Agent on the remaining GPU-equipped nodes. Users view the dashboard provided by the GPU-equipped node with AIBooster Server installed from their personal PCs through a browser via TCP port 3000.

Setup Procedure​

AIBooster Server​

Copy and execute the following command in your terminal on the management node where AIBooster Server will run:

curl -LO https://assets.aibooster.fixstars.com/faibup-server.sh && sudo sh ./faibup-server.sh

As described in the Troubleshooting page, some ports on the management node running AIBooster Server must be open/accessible.

AIBooster Agent​

On the node where AIBooster Agent will run, replace [Management Node IP Address] in the following command with the actual IP address, then execute it:

curl -LO https://assets.aibooster.fixstars.com/faibup-agent.sh && sudo sh ./faibup-agent.sh -- --server_endpoint http://[Management Node IP Address]

Dashboard Initial Configuration​

Grafana First Login​

AIBooster uses Grafana for performance data visualization. When accessing the dashboard from a browser for the first time, the following screen will be displayed:

Grafana-login

Password Configuration​

  1. Enter admin as both username and initial password
  2. When prompted to change the administrator password, enter any password

Configuration Complete Confirmation​

When the following screen is displayed, the initial dashboard configuration is complete:

Grafana-top