NVIDIA Base Command Manager 11 - User Manual - Revision d7928f6

Oct 22, 2026

NVIDIA Base Command Manager 11 user manual describing cluster access and usage workflows, from logging in and setting up user environments to loading environment modules and compiling applications. It explains MPI basics and practical build and run steps, including Open MPI, module path selection, interconnect and compiler choices, and hybrid execution recommendations. It also covers workload management concepts with job submission and Slurm usage, GPU programming paths (CUDA and OpenCL), Kubernetes operations and privileges, Spark-on-Kubernetes execution, and Jupyter notebook kernel provisioning.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages105

Reading Time4 h 25 min

AccessFree to download

0.1 About This Manual

0.2 Getting User-Level Support

1 Introduction

1.1 What Is A Beowulf Cluster?

1.2 Brief Network Description

2 Cluster Usage

2.1 Login To The Cluster Environment

2.2 Setting Up The User Environment

2.3 Environment Modules

2.4 Compiling Applications

3 Using MPI

3.1 Introduction

3.2 MPI Libraries

3.3 MPI Packages And Module Paths

3.5 Compiling And Carrying Out An MPI Run

4 Workload Management

4.1 What Is A Workload Manager?

4.4 Job Submission Process

5 Slurm

5.1 Loading Slurm Modules And Compiling The Executable

5.2 Running The Executable With salloc

5.3 Running The Executable As A Slurm Job Script

6 Using GPUs

6.2 Using CUDA

6.3 Using OpenCL

6.5 Available Tools

7 Using Kubernetes

7.1 Introduction To Kubernetes Running Via NVIDIA Base Command Manager

7.3 Kubernetes Quickstarts

8 Spark On Kubernetes

8.2 Running Spark Jobs Via The Kubernetes Spark Operator

8.3 Running Spark Jobs Directly Via spark-submit

9 User Portal

9.1 Overview Page

9.6 Accounting And Reporting Mode

10 Using Jupyter

10.2 Jupyter Notebook Examples

10.3 Jupyter Kernels

NVIDIA Base Command Manager 11 User Manual Revision: d7928f6 Date: Wed Oct 21 2026 Table of Contents Table of Contents ............................................. 3 0.1 About This Manual ......................................... 7 0.2 Getting User-Level Support .................................... 7 1 Introduction 9 1.1 What Is A Beowulf Cluster? .................................... 9 1.1.1 Background And History ................................. 9 1.1.2 Brief Hardware And Software Description ....................... 9 1.2 Brief Network Description ..................................... 10 2 Cluster Usage 13 2.1 Login To The Cluster Environment ................................ 13 2.2 Setting Up The User Environment ................................ 14 2.3 Environment Modules ....................................... 14 2.3.1 Available Commands ................................... 14 2.3.2 Managing Environment Modules As A User ...................... 16 2.3.3 Changing The Default Environment Modules ..................... 17 2.4 Compiling Applications ...................................... 18 2.4.1 Open MPI And Mixing Compilers ............................ 19 3 Using MPI 21 3.1 Introduction ............................................. 21 3.2 MPI Libraries ............................................ 21 3.3 MPI Packages And Module Paths ................................. 21 3.3.1 MPI Packages That Can Be Installed, And Their Corresponding Module Paths .. 22 3.3.2 Finding The Installed MPI Packages And Their Available Module Paths ...... 23 3.4 The Appropriate Interconnect, Compiler, And MPI Implementation For A Module .... 24 3.4.1 Interconnects ........................................ 24 3.4.2 Selecting A Compiler And MPI implementation .................... 24 3.5 Compiling And Carrying Out An MPI Run ........................... 24 3.5.1 Example MPI Run ..................................... 25 3.5.2 Hybridization ........................................ 30 3.5.3 Support Thread Levels ................................... 32 3.5.4 Further Recommendations ................................ 32 4 Workload Management 33 4.1 What Is A Workload Manager? .................................. 33 4.2 Why Use A Workload Manager? ................................. 33 4.3 How Does A Workload Manager Function? ........................... 33 4.4 Job Submission Process ....................................... 34 4.5 What Do Job Scripts Look Like? .................................. 34 4.6 Running Jobs On A Workload Manager ............................. 34 5 Slurm 35 5.1 Loading Slurm Modules And Compiling The Executable ................... 35 5.2 Running The Executable With salloc .............................. 36 5.2.1 Node Allocation Examples ................................ 36 5.3 Running The Executable As A Slurm Job Script ......................... 38 5.3.1 Slurm Job Script Structure ................................. 38 5.3.2 Slurm Job Script Options ................................. 39 5.3.3 Slurm Environment Variables ............................... 39 5.3.4 Submitting The Slurm Job Script With sbatch ..................... 40 5.3.5 Checking And Changing Queued Job Status With squeue, scancel, scontrol And sview ............................................. 40 6 Using GPUs 43 6.1 Packages ............................................... 43 6.2 Using CUDA ............................................. 43 6.3 Using OpenCL ............................................ 44 6.4 Compiling Code ........................................... 44 6.5 Available Tools ........................................... 45 6.5.1 CUDA gdb ......................................... 45 6.5.2 The nvidia-smi Utility .................................. 45 6.5.3 CUDA Utility Library ................................... 46 6.5.4 CUDA “Hello world” Example ............

FAQs

How do users log in and set up the user environment in the cluster?

The manual provides steps for logging in to the cluster environment and configuring the user environment before using modules and running workloads.

What guidance is given for compiling and running MPI applications?

It covers selecting MPI libraries and module paths, choosing the appropriate interconnect, compiler, and MPI implementation, and then compiling and performing an MPI run with examples and recommendations for hybridization and thread levels.

How does the manual explain workload management and Slurm job submission?

It introduces workload manager concepts, describes the job submission process and job script structure, and then details Slurm module loading, execution via salloc, batch scripts for sbatch, and job status checks using squeue, scancel, scontrol, and sview.

How are Kubernetes, Spark-on-Kubernetes, and Jupyter integrated for users?

The manual explains Kubernetes privileges and quickstarts with kubectl and Helm, then covers Spark jobs via the Spark operator or spark-submit, and concludes with Jupyter notebook usage including kernel provisioning and kernel creator extensions.

Oct 22, 2026

Upload by How to eat