Introduction

This page is for version 11 (default). For other versions, use one of these documentation indexes: - 11 (default): https://docs.nvidia.com/base-command-manager/product-manuals/11/llms.txt - 10: https://docs.nvidia.com/base-command-manager/product-manuals/10/llms.txt

In weather, a cloudburst is used to convey the idea that a sudden flood of cloud contents takes place. In cluster computing, the term cloudbursting conveys the idea that a flood of extra cluster capacity is made available when needed from a cloud computing services provider such as Amazon.

NVIDIA Base Command Manager implements cloudbursting for two scenarios:

1. A “Cluster On Demand”, or a “pure” cloud cluster (Chapter 2). In this scenario, the entire cluster can be started up on demand from a state of non-existence. All nodes, including the head node, are instances running in a coordinated manner entirely inside the cloud computing service.

2. A “Cluster Extension”, or a “hybrid” cloud cluster (Chapter 3). In this scenario, the head node is kept outside the cloud. Zero or more regular nodes are also run outside the cloud. When additional capacity is required, the cluster is extended via cloudbursting to make additional nodes available from within the cloud.

Chapters 2 and 3 deal with mainly the GUI configuration of the Cluster On Demand and Cluster Extension scenarios.

Chapter 4 looks at mainly command line tools for configuration of the Cluster On Demand and Cluster Extension scenarios, considering mainly AWS.

Chapter 5 looks at Cluster Extension for Azure.

Chapter 6 discusses some miscellaneous aspects of cloudbursting.