Introduction | NVIDIA Base Command Manager

1.1 NVIDIA Base Command Manager Functions And Aims

NVIDIA Base Command Manager (often shortened to BCM) contains tools and applications to facilitate the installation, administration, and monitoring of a cluster. In addition, BCM aims to provide users with an optimal environment for developing and running applications that require extensive computational resources.

1.2 The Scope Of The Administrator Manual (This Manual)

The Administrator Manual covers installation, configuration, management, and monitoring of BCM, along with relevant background information to help understand the topics covered.

1.2.1 Installation

Installation can generally be divided into parts as follows, with some parts covered by the Administrator Manual, some by the Installation Manual, and some by other manuals:

Initial installation of BCM: This is covered in the Installation Manual, which gives a short introduction to the concept of a cluster along with details on installing BCM onto the head node. The Installation Manual is therefore the first manual an administrator should usually turn to when getting to work with BCM for the first time. The Administrator Manual can be referred to as the main reference resource once the head node has had BCM installed on it.
Provisioning installation: This is covered in the Administrator Manual. After the head node has had BCM installed on it, the other, regular, nodes can (network) boot off it and provision themselves from it with a default image, without requiring a Linux distribution DVD themselves. The network boot and provisioning process for the regular nodes is described in detail in Chapter 5.

In brief, provisioning installs an operating system and files on a node. This kind of installation to a regular node differs from a normal Linux installation in several ways. An important difference is that content that is put on the filesystem of the regular node is normally overwritten by provisioning when the regular node reboots.
Post-installation software installation: The installation of software to a cluster that is already configured and running BCM is described in detail in Chapter 11 of this manual.
Third-party software installation: The installation of software that is not developed as part of BCM, but is supported as a part of BCM. This is described in detail in the Installation Manual.
Cloudbursting, Edge, and Machine Learning: these are integrated as part of BCM in various ways. These have their own deployment procedures and have separate manuals.

1.2.2 Configuration, Management, And Monitoring Via BCM Tools And Applications

The administrator normally deals with the cluster software configuration via a front end to BCM. This can be GUI-based (Base View, section 2.4) or shell-based (cmsh, section 2.5). Other tasks can be handled via special tools provided with BCM, or the usual Linux tools. The use of BCM tools is usually recommended over standard Linux tools because cluster administration often has special issues, including that of scale.

The following topics are among those covered in this manual:

Chapter	Title	Description
2	Cluster Management With NVIDIA Base Command Manager	Introduction to main concepts and tools of BCM. Lays down groundwork for the remaining chapters
3	Configuring The Cluster	Further configuration and set up of the cluster after software installation of BCM on the head node.
4	Power Management	How power management within the cluster works
5	Node Provisioning	Node provisioning in detail
6	User Management	Account management for users and groups
7	Workload Management	Workload management implementation and use
8	The `cm-scale` Service	A BCM service to dynamically scale the cluster according to need
9	Ceph Installation	Installing Ceph scalable storage
10	BeeGFS Installation	Installing BeeGFS high performance storage
11	Post-Installation Software Management	Managing, updating, modifying BCM software and images
12	Monitoring: Monitoring Cluster Devices	Device monitoring and conditional action triggers
13	Monitoring: Job Monitoring	Jobs resource consumption monitoring by the jobs
14	Monitoring: Job Accounting	Jobs resource consumption monitoring aggregated by user or similar groupings
15	Monitoring: Job Chargeback	Resource request monitoring, so that groups of users can be charged for their use
16	Day-To-Day Administration	Miscellaneous administration
17	High Availability	Background details and setup instructions to build a cluster with redundant head nodes
18	The Jupyter Notebook Environment Integration	Installing and using the Jupyter notebook environment

The appendices to this manual generally give supplementary details to the main text.

The following topics are also logically a part of BCM administration, but they have their own separate manuals. This is because they have, or are eventually expected to have, many features:

Cloudbursting (Cloudbursting Manual)
Machine Learning (Machine Learning Manual)
Edge deployment (Edge Manual)
Developer topics (Developer Manual)
Containerization topics (Containerization Manual)

1.3 Outside The Direct Scope Of The Administrator Manual

The following supplementary resources can deal with issues related to this manual, but are outside its direct scope:

Use by the end user: This is covered very peripherally in this manual. The user normally interacts with the cluster by logging into a custom Linux user environment to run jobs. Details on running jobs from the perspective of the user are given in the User Manual.
The NVIDIA Enterprise Support Portal at https://enterprise-support.nvidia.com/ carries Knowledge Articles that often supplement the Administrator Manual with discussion of the following:
- Obscure, or complicated, configuration cases
- Procedures that are not really within the scope of BCM itself, but that may come up as part of related general Linux configuration.
Further support options. If the issue is not described adequately in this manual, then section 16.2 describes how to get further support.

Download this manual as a PDF