Introduction

This document walks you through the process of accessing and using Microway's Test Drive Cluster. This cluster is an HPC resource which we make available for benchmarking and development projects.

Available Resources

Each Compute Node in the Test Drive Cluster contains:

• Latest-generation multicore CPUs:

  • Dual 16-core 2.8GHz Intel Xeon 6242 CPUs

  • Dual 8-core, 16-core, 32-core, and 64-core AMD EPYC "Rome" CPUs

• 256GB to 768GB DDR4 Memory

• 100Gbps EDR and 200Gbps HDR InfiniBand interconnects

Additionally, users may access the following GPU compute accelerators:

• NVIDIA Tesla P100 PCI-E 16GB GPU

• NVIDIA Tesla V100S PCI-E 32GB GPU

• NVIDIA A100 PCI-E 40GB GPUs with NVLink pairs (coming soon)

• NVIDIA DGX A100 (coming soon)

Quick Start

In brief, here are the steps to running on the Test Drive Cluster: 1. Request an account on the cluster 2. Log into the cluster via encrypted connection 3. Upload your code and input data to the cluster 4. Compile your code (if necessary) 5. Run your code on a Compute Node

This manual walks you through the process.

Assumptions and Disclaimers

• Being handed the keys to a race car does not guarantee you know how to drive a race car. You may need practice...

  • Using an HPC system means using Linux. If you are not already familiar with Linux, you will need to learn some of the basics.

  • Fully utilizing HPC resources often requires some understanding of the hardware upon which you are running.

• We will cover many topics, but we cannot encompass every issue and pitfall.