Overview

The CHAI Cluster Manager Tool is a unified, modular automation framework designed to simplify and accelerate the deployment, configuration, and management of HPC and AI clusters. It integrates provisioning, orchestration, workload scheduling, and monitoring using industry-standard tools like xCAT, Ansible, SLURM, OpenLDAP and Kubernetes etc.

The framework currently supports:

  • x86_64 architecture and multi OS-based environments

  • Multiple software stack versions with version control

  • Bare Metal and Virtual Machine

  • Both bare-metal and containerized software stack of HPC-AI infrastructure

  • Multi-tenant Kubernetes control planes for AI user(Team) isolation

HPC Cluster Management and Service Nodes

To efficiently manage the cluster, dedicated service nodes handle installation, deployment, and administrative tasks. These include:

  • Head Node - A head node which is CHAI Manger tool. To configure all the service nodes. (It can be temporary or permanent.)

  • Master Nodes – Manage compute nodes using xCAT for provisioning and SLURM for workload scheduling and ldap for central user authentication.

  • AI Nodes – Kubernetes based orchestration for AI clusters, supporting containerized AI workloads and workload management.

  • Management Nodes – Handle monitoring, logging, and ticketing systems.

  • Login Nodes – Provide user access to the cluster.

  • BMC (Baseboard Management Controller) Nodes – Oversee hardware health and remote management.

  • Firewall Node – Ensures network security