NVIDIA Base Command Manager (BCM) provides cluster management software for streamlining cluster provisioning, workload management, and infrastructure monitoring. It provides all the tools needed for deploying and managing an AI data center.

Base Command Manager is included in NVIDIA AI Enterprise, which brings together optimized micro services, frameworks, and libraries for AI development, along with advanced GPU orchestration and infrastructure management, into a production-grade software suite for your business.

Supported Linux Distributions

Base Command Manager 11 is available for the following Linux distributions:

  • Rocky Linux 8/9
  • SLES 15
  • Ubuntu 22.04/24/26

You can specify your desired Linux distribution on the ISO download page. The selected distribution is pre-packaged within the resulting ISO file.

BCM is typically installed on a head node, or on a pair of head nodes in a high-availability configuration. Images are provisioned to the non-head nodes from the head node(s). By default, the OS used for the images matches that of the head node(s), but this can be changed later.

Cluster Structure

In its most basic form, a cluster running BCM contains:

  • One machine designated as the head node
  • Several machines designated as compute nodes
  • One or more (possibly managed) Ethernet switches
  • One or more power distribution units (optional)

The head node is the most important machine in the cluster because it controls all other devices, such as compute nodes, switches, and power distribution units. It is also the host that all users, including the administrator, log in to by default.

Minimum Hardware Requirements

The following are the minimum hardware requirements, suitable for a cluster of one head node and two regular compute nodes:

Head Node

  • x86-64 or ARMv8 CPU
  • For running the application software:
    4GB RAM for the x86 architecture

    16GB RAM for the ARMv8 architecture
  • 80GB of disk space
  • 1 Gigabit Ethernet NICs (2 NICS for the most common Type 1 topology)
  • DVD drive or USB drive

Compute Nodes

  • x86-64 or ARMv8 CPU
  • 1GB RAM (at least 4GB is recommended for diskless nodes)
  • 1 Gigabit Ethernet NIC

Supported Hardware

An NVIDIA-certified system for AI Enterprise is a system with NVIDIA GPUs and networking that has passed a set of certification tests validating its performance, reliability, and scale for data centers. Many system builders partnered with NVIDIA provide such certified hardware; these builders are listed in NVIDIA’s qualified system catalog.

The following hardware is also supported, even if it may not be NVIDIA-certified for AI Enterprise:

Compute Nodes

  • ASUS, Atos, Cisco, Dell EMC, Fujitsu, HPE (Hewlett Packard Enterprise),IBM,Lenovo,NVIDIA DGX,Oracle, Supermicro

Other brands are also expected to work, even if not explicitly supported.

Ethernet Switches

  • Cisco, Dell,HP Networking Switches,Netgear,Supermicro

Other brands are also expected to work, although not explicitly supported.

Power Distribution Units

  • APC (American Power Conversion) Switched Rack PDU

Other brands with the same SNMP MIB mappings are also expected to work, although not explicitly supported.

Management Controllers

  • HP iLO
  • iDRAC
  • IPMI 1.5/2.0
  • Redfish v1

InfiniBand and Similar

  • NVIDIA (formerly Mellanox) HCAs, and most other InfiniBand HCA
  • NVIDIA (formerly Mellanox), and most other InfiniBand switches

GPUs

  • NVIDIA GPUs later than Volta, with the latest recommended drivers
  • NVIDIA DGX OS 7
  • NVIDIA DGX OS 8, x86_64 systems only (B300, B200, H200, H100, H800, A100, A800). As of DGX OS 8.0, ARM64 systems, including GB200 and GB300, are not supported
  • AMD Radeon GPUs, as listed on AMD’s Radeon software release notes page

GPUs that are NVIDIA-certified for AI Enterprise are listed in the product support matrix on NVIDIA’s documentation site.

RAID

Software or hardware RAID is supported. Fake RAID is not considered a serious production option and is supported accordingly.

Downloading the ISO

You can download the installer ISO using the link below:

https://customer.brightcomputing.com/download-iso

Step-by-Step Deployment

Step 1: Boot the Installer

Select “Start Base Command Manager Graphical Installer,” the recommended option, to begin installation.

Step 2: Start the Installation

Click “Start Installation” to begin.

Step 3: Accept the License

Accept the license agreement and click Next.

Step 4: Kernel Modules

The Kernel Modules screen shows the kernel modules recommended for loading, based on a hardware probe. You can add or remove modules according to your requirements.

Step 5: Hardware Info

The Hardware Info screen gives an overview of the hardware detected by the probe used to load kernel modules. If any hardware is not detected at this stage, use the Back button to return to the Kernel Modules screen, add the appropriate modules, and then return to the Hardware Info screen to confirm the hardware is detected.

Step 6: Installation Source and Media Integrity Check

Clicking Next starts the media integrity check. Once the check completes, click Next to continue.

Step 7: Cluster Settings

The following settings can be configured on this screen:

  • Cluster name
  • Administrator email — where mail to the administrator is sent; this does not need to be a local address
  • Time zone
  • Time servers — the defaults are pool.ntp.org servers; a time server is recommended to avoid time discrepancies between nodes
  • Nameservers — one or more DNS servers can be set
  • Search domains — one or more domains can be set
  • Environment modules — traditional Tcl modules are set by default; Lmod is an alternative

Step 8: Workload Manager

The Workload Manager screen lets you select a supported workload manager. If none is selected here, one can be installed later, after the cluster has been deployed.

Step 9: Network Topology

The Network Topology screen lets you choose one of three network topologies :

  • Type 1 — the most common and simplest option. The head node provides DHCP and PXE services (during the pre-init boot stage only) to a secondary, isolated network for the worker nodes, keeping cluster traffic segregated. The external, typically corporate, network is used only to reach the head node(s) for management.
  • Type 2 — nodes are connected via a router to a public network. A packet from a regular node destined for outside the cluster goes through the router rather than the head node, while traffic between regular nodes and the head node still goes directly, since they normally share the same network.
  • Type 3 — nodes are connected on a routed public network. A packet from a regular node destined for another network uses a router, and the head node, being on a different network, can also only be reached via a router. By default, the regular nodes’ network is called Internal net and the head node’s network is called Management net. Any routing beyond these networks is configured on the routers, not on the cluster.

I chose the Type 2 network for this deployment; choose the topology that best fits your requirements.

Step 10: Head Node

Enter the hostname and password for the head node.

Step 11: Compute Nodes

The Compute Nodes screen lets you set:

  • The number of racks used to host the nodes
  • The number of nodes
  • The naming format for the nodes, made up of:
    • Node base name (default: node)
    • Node start number (default: 1)
    • Node digits (default: 3)

By default, the first compute node is named node001, the second node002, and so on.

Step 12: BMC Configuration

Configure BMCs compatible with IPMI, iDRAC, iLO, or CIMC. BMCs can be configured for the head node or the compute nodes. I chose to add servers manually after deployment.

Step 13: Networks

The available options here depend on the network topology chosen earlier and show the IP configuration required. I selected a single network, based on the Type 2 topology chosen previously.

Step 14: Head Node Interfaces

Assign an IP address to the head node.

Step 15: Compute Node Interfaces

The boot interface (BOOTIF) is used to pick up the image for a node during provisioning. The IP offset is used to calculate the IP address assigned to a regular node’s interface. Since nodes are numbered sequentially, their interfaces are typically also given network IP addresses in sequence on the selected network — by default, BCM assigns interface IP addresses sequentially, in steps of 1, starting after the network’s base address. The default IP offset is 0.0.0.0, meaning node interfaces start their range at the network’s usual default values.

Step 16: Disk Layout

Select the drive on the head node where the BCM software will be installed.

Step 17: Disk Layout Settings

Set the disk partitioning layout for the head node and compute nodes, using the two available options:

  • Head node disk layout — by default, one big partition if the drive is smaller than about 500GB, or several partitions if the drive is 500GB or larger
  • Compute nodes disk layout — by default, several partitions, using the Default Standard Layout

A partitioning layout other than the default can be selected from the drop-down options for the head node and regular nodes. Available options include RAID, failover, STIG-compliant, and LUKS schemes. The add icon lets you supply a custom partitioning layout, either as a file or from the default template as a starting point, and the edit icon lets you modify an existing layout.

Step 18: Additional Software

This screen lists additional software that can be added to the cluster, provided it was included when the installer ISO was generated. CUDA is one example of additional software. If you plan to use NVIDIA AI Enterprise, leave these options unselected.

Step 19: Summary

Verify the configuration details, then click Start to begin the installation.

Step 20: Reboot

Once deployment is complete, you will be prompted to reboot. Reboot the server.

Step 21: Access the BCM UI

Access the BCM UI using the provided URL, and log in.

In my next post, I’ll cover how to add nodes and run jobs, and how to deploy a Kubernetes cluster with AI operators.