BCM architecture and concepts

NVIDIA Base Command Manager (BCM) (formerly Bright Cluster Manager) is the comprehensive cluster management platform for HPC and AI clusters.

BCM handles:

• Node provisioning — PXE boot, OS imaging, driver installation
• Cluster configuration — Networking, storage, services
• Workload scheduling — Slurm and Kubernetes integration
• Monitoring — Node health, GPU status via DCGM/NVSM
• Software deployment — Automated distribution

Architecture:

• Head Node — Runs BCM software, provides DHCP/PXE, NFS
• Compute Nodes — Provisioned and managed by BCM
• BMC/iDRAC/iLO — Out-of-band hardware management

cmsh is the BCM command shell:

cmsh                               # Start BCM shell
cmsh> device list                  # List all cluster devices
cmsh> device status <node>         # Node status
cmsh> device get <node>            # Node configuration
cmsh> softwareimage list           # List available OS images
cmsh> services list                # BCM services status