Cluster Planning

InfiniBand Cluster Bring-up Procedure

Effective cluster planning lays the groundwork for an efficient system and good monitoring and debugging ability.

The planning procedures in this document are arranged in a sequential order for a new cluster installation. If you are not installing a new cluster, you might need to choose which procedures to use. However, you should still perform them in the order they appear in the Cluster planning section.

To plan your cluster, complete the following procedures:

  1. Set the Topology type. For information on how to do so, see Setting the InfiniBand Cluster Topology.

  2. Create a Point-to-Point excel file. For information on how to do so, see Creating a Point-to-Point Excel File.

  3. Create a Topology file. For information on how to do so, see Creating a Topology File.

  4. Save the Topology file for future usage. For information on how to do so, see Saving the Topology File.

  5. Confirm the Topology was created and set as desired. For information on how to do so, see Confirming Topology.

When you are ready to install the components with which you plan to build your cluster, Confirm Components' Firmware and Software Versions. If update is needed to one of the components, review the information in section On-site Upgrade – Low scale to ensure that you have the latest information.

© Copyright 2024, NVIDIA. Last updated on May 28, 2024.