This post is authored by Alex Johnson, Principal HPC Partner Solutions Architect at AWS, and Jamie Lee, Senior Principal Engineer at Ansys, along with Taylor Smith, Director of the Fluids Business Unit at Ansys.
Today, we will explore the performance and related costs of executing computational fluid dynamics (CFD) simulations on AWS with Ansys Fluent through the Ansys Gateway powered by AWS, which we will refer to as “Ansys Gateway” throughout this article.
Ansys Gateway is a solution hosted on the AWS Marketplace that allows users to manage their entire Ansys virtual desktop and HPC simulation workflows, featuring over fifty Ansys products within the AWS cloud ecosystem. Notable organizations such as Emirates Team New Zealand and Turntide Technologies utilize Ansys Gateway to enhance their design and engineering simulation processes.
Ansys Fluent is an advanced simulation software widely employed by engineers and scientists in sectors like automotive, aerospace, manufacturing, and energy to drive innovation and improve product development.
In this analysis, we will assess the performance and pricing for three distinct test cases across various configurations of Amazon Elastic Compute Cloud (Amazon EC2) instance types. The insights gleaned from this post will assist you in selecting the right hardware for running Ansys Fluent simulations.
Overview of Ansys Gateway
Ansys Gateway serves as a secure, online platform that empowers users to create, manage, and execute comprehensive computer-aided engineering (CAE) workflows within their AWS accounts. Earlier this year, we published a blog that detailed the solution architecture components, security measures, and standard user workflows for utilizing Ansys Gateway.
The Ansys applications (solvers) accessible via Ansys Gateway are pre-configured, validated, and rigorously benchmarked against various Amazon EC2 hardware for both performance and cost. Users can refer to the recommended usage guidelines in the Ansys Gateway documentation within Ansys Help to set up their virtual desktop infrastructure (VDI) or HPC environment using the suggested Amazon EC2 instance types. This enables users to conduct simulations with their chosen solvers right away, without needing to manually configure and optimize the solvers, simulation settings, and hardware parameters.
Benchmark Details and Simulation Environment
For benchmarking purposes, we utilized test cases from the standard Ansys Fluent Benchmarks suite. The model descriptions for each test case, including mesh size represented by the number of cells, turbulence model utilized, and fluid-flow conditions, are summarized in Table 1. These benchmarks (illustrated in Figures 1a – 1c) reflect the typical size and physics modeled by users. We employed Ansys Fluent version 2023 R1 for running the simulations.
Table 1: Summary of Ansys Fluent Benchmark Test Cases
Figure 1: Visual representations of Ansys Fluent benchmarking test cases:
- a) Steady-state simulation of flow through a vehicle exhaust system with 33 million cells;
- b) Transient simulation of flow through a combustor with 71 million cells;
- c) Steady-state simulation of external aerodynamics of a Formula 1 race car with 140 million cells.
Simulation Environment
AWS has recently introduced the Amazon EC2 Hpc7a instance type, which is powered by 4th generation AMD EPYCTM (Genoa) processors featuring up to 192 physical cores and 300 Gbps Elastic Fabric Adapter (EFA) network bandwidth. We assessed the performance of these benchmarks on both Hpc7a and the previous generation Hpc6a instances.
Table 2: Amazon EC2 Instances Used for Benchmark Testing
Analysis Methodology
Our goal in conducting these benchmarks was to quantify Ansys Fluent performance, associated hardware, platform, and licensing expenses, enabling us to recommend the most suitable Amazon EC2 instance type for use on Ansys Gateway. We analyzed the following metrics:
- Solver Rating: This metric indicates solver performance by calculating how many times the benchmark can be executed on a specific machine within 24 hours. A higher Solver Rating reveals improved performance. We ran simulations for 1000 iterations for steady-state flow or 1000 timesteps for transient flow.
- Job Cost: The total job (or simulation) cost consists of three main components: the Amazon EC2 cost, Ansys Gateway charge of $0.25 per running Amazon EC2 instance per hour, and the Ansys licensing cost.
These cost representations will help you select the right HPC configuration to achieve one of three simulation objectives: minimize job cost, maximize performance, or obtain the best performance/cost ratio.
For this analysis, we utilized Ansys Elastic Licensing to represent licensing costs, which may not fully apply if you use leased or perpetual licenses. Moreover, we did not factor in storage and networking costs, primarily because compute expenses dominate the infrastructure costs for these simulations. The Amazon EC2 on-demand pricing from the us-east-2 (Ohio) region was used. You can also explore flexible pricing options like Compute Savings Plans or Reserved Instances (RI) that offer significant savings up to 72% compared to On-Demand pricing.
Results
To gauge Ansys Fluent performance, we charted the Solver Rating against the number of instance cores for each of our three test cases, displayed in Figures 2 (a, b, c). A higher Solver Rating signifies superior performance. The Vehicle Exhaust model, featuring 33 million cells, scaled to 1536 cores, while the Combustor and F1 Race Car models scaled to over 6000 cores each.
Figure 2a: Solver Rating comparison for the Vehicle Exhaust benchmark across Amazon EC2 Hpc6a and Hpc7a instance cores.
Figure 2b: Solver Rating comparison for the Combustor benchmark across Amazon EC2 Hpc6a and Hpc7a instance cores.
Figure 2c: Solver Rating comparison for the Formula 1 car benchmark across Amazon EC2 Hpc6a and Hpc7a instance cores.
Due to the differing physical core counts between Hpc6a and Hpc7a instance types, the relative performance between the two varies at both the core and node levels. Figures 3 (a, b, c) illustrate the Solver Rating variation with the number of instances (nodes).
Figure 3a: Solver Rating comparison for the Vehicle Exhaust benchmark across Amazon EC2 Hpc6a and Hpc7a instances (nodes).
Figure 3b: Solver Rating comparison for the Combustor benchmark across Amazon EC2 Hpc6a and Hpc7a instances (nodes).
For further details on related topics, check out this informative blog post here. You can also find rich insights from this source, an authority on this subject. Additionally, if you’re curious about life as an employee, this article on Quora provides excellent insight.
For location reference, Amazon IXD – VGT2 is situated at 6401 E Howdy Wells Ave, Las Vegas, NV 89115.
Leave a Reply