/of-on-aws-tests

Six OpenFOAM test cases - used to gauge the performance of EC2 instances running CFD (and provide comparison to your local machines)

Primary LanguageC++GNU General Public License v3.0GPL-3.0

Choosing an AWS EC2 Instance for OpenFOAM - Test Cases

These cases are supporting material for an article that looked at how to choose an EC2 instance to run OpenFOAM (computational fluid dynamics) simulations.

The article summarises which instances were tested and how it was done, along with a rough guide to choosing an EC2 instance for your own OpenFOAM simulations.

This repository contains the test cases used to prepare that article. It contains 6 versions of the windAroundBuildings tutorial from OpenFOAM® v6. Each with a different blockMesh resolution to produce a different model size.

Test Case Model Size blockMesh
x1 (STD) ~185K Cells 25 - 20 - 10
x2 ~1.1 Million Cells 50 - 40 - 20
x3 ~3.3 Million Cells 75 - 60 -30
x4 ~7.0 Million Cells 100 - 80 - 40
x5 ~13.3 Million Cells 125 - 100 - 50
x6 ~22.5 Million Cells 150 - 120 - 60

Grabbing these cases & running them locally will give you some additional data points to compare the performance of your local machines against some current EC2 instances. It should also give you an idea of how your usual simulations might perform on AWS.

Prerequisites

The test cases in the article were run in OpenFOAM v6 on Ubuntu 16.04 - they will (most likely) run in earlier versions of OpenFOAM, but may require modification. They should work out-of-the-box in v6.

Instructions for use

  1. Download the cases to your test machine with git OR download & extract the zip archive
    • git clone https://github.com/CFDEngine/of-on-aws-tests.git
    • wget -O of-on-aws-tests.zip https://github.com/CFDEngine/of-on-aws-tests/archive/master.zip
  2. Choose which test case is closest to your usual model size & change into that directory
  3. Change the numSubDomains in decomposeParDict to suit the number of physical cores you have available
  4. Run & time the case using the Allrun file - on Ubuntu this can be done with the following command
    • /usr/bin/time -o log.time ./Allrun

The time command will save it's output into a log.time file – the contents of which should look something like this:

61.85user 0.60system 1:03.92elapsed 97%CPU (0avgtext+0avgdata 362380maxresident)k
10416inputs+0outputs (4major+103628minor)pagefaults 0swaps

In this example 1:03.92elapsed is the interesting bit, showing the time taken was 1m 3.92s.

See the full article for links to the complete data set of equivalent timings on AWS.