Benchmarking & Scaling Tutorial

From HPC Wiki
Jump to navigation Jump to search


Tutorial
Title: Benchmarking & Scaling
Provider: HPC.NRW

Contact: tutorials@hpc.nrw
Type: Online
Topic Area: Performance Analysis
License: CC-BY-SA
Syllabus

1. Introduction & Theory
2. Interactive Manual Benchmarking
3. Automated Benchmarking using a Job Script
4. Automated Benchmarking using JUBE
5. Plotting & Interpreting Results


This tutorial is meant to help you benchmark your application on an HPC cluster environment and to optimise the use of available resources. We will explain why you should benchmark a software you plan to run on a cluster before doing any productive calculations, what scalability is, tell you about potential bottlenecks and refer to advanced toolkits which can automate the process.



Part I

Introduction and Theory

Interactive Manual Benchmarking

Automated Benchmarking using a Job Script

Automated Benchmarking using JUBE

Plotting and Interpreting Results