SU2 and the Challenge of Scalability: How CloudHPC is Speeding Up CFD Simulations
Engineers are constantly pushing the boundaries of what’s possible, and that often means tackling complex simulations that demand immense computational power. In the world of Computational Fluid Dynamics (CFD), one of the most powerful tools available is the SU2 software suite. But even with state-of-the-art software, a fundamental challenge remains: scalability.
What is SU2?
SU2 is a suite of open-source software tools written in C++ for the numerical solution of partial differential equations (PDEs). While its primary applications are CFD and aerodynamic shape optimization, its capabilities have been extended to solve more general equations like electrodynamics and chemically reacting flows. A key feature of SU2 is its ability to perform PDE-constrained optimization. It also supports both continuous and discrete adjoint methods for calculating sensitivities and gradients of a scalar field.
Essentially, SU2 is an open platform for engineers and researchers to prototype new numerical methods, discretization schemes, and parallelization techniques. Because of its open-source nature and extensibility, it’s a popular choice for a wide range of applications in the aeronautical, automotive, and renewable energy industries, among others.
Why is Scalability So Difficult?
Scalability is the ability to reduce the time needed to get a solution by increasing the number of hardware resources used. The goal is simple: more cores should equal faster results. In a perfect world, if you double your CPU count, you’d cut your simulation time in half. But in reality, this is incredibly difficult to achieve.
The relationship between adding more CPUs and reducing simulation time, known as SpeedUp (S=T1/TN), is not linear. After a certain point, increasing the number of CPUs can provide a low-efficiency speedup, no speedup at all, or even reduce the speedup.
This limitation, or “scalability limit,” depends on several factors:
- Software Version: The efficiency of a particular software version.
- Hardware Used: The specific processors and memory architecture.
- Simulation Type: The complexity of the models involved.
- Mesh: The size and structure of the computational mesh.
The graph above shows how the speedup increases with the number of processors (Np), but eventually plateaus, demonstrating that simply adding more CPUs isn’t a magic bullet. The challenge lies in finding the optimal balance and the right hardware to push that limit as far as possible.
CloudHPC and High-Performance Hardware 🚀
This is where CloudHPC comes in. CloudHPC, developed by CFD FEA SERVICE, is a cloud computing cluster specifically designed for high-CPU-demanding engineering applications like CFD. It’s accessible via a web browser, making it easy for engineers to run simulations without the hassle of configuring, installing, or setting up complex systems.
CloudHPC has made SU2 version 8.3.0 available on its platform. This is a huge advantage, as users can directly benefit from the performance improvements and bug fixes of the latest version.
Furthermore, CloudHPC leverages powerful hardware to overcome scalability challenges. It offers access to AMD EPYC™ MILAN and Intel® Xeon® Emerald Rapids processors, which are specifically designed for high-performance computing. The results from the presentation show how different processor types and their architectures, like the large Last Level Cache (LLC) on the Intel processor, can significantly affect scalability.
By using CloudHPC’s platform, you can speed up a single simulation by allocating more computing power, a concept known as vertical scalability. The platform allows for up to 224 vCPUs or 192 cores for a single simulation. This, combined with the power of modern AMD and Intel processors, provides a robust solution for running large-scale CFD simulations efficiently. With CloudHPC, engineers can focus on their designs and analyses, leaving the hardware and scalability headaches behind.
CloudHPC is a HPC provider to run engineering simulations on the cloud. CloudHPC provides from 1 to 224 vCPUs for each process in several configuration of HPC infrastructure - both multi-thread and multi-core. Current software ranges includes several CAE, CFD, FEA, FEM software among which OpenFOAM, FDS, Blender and several others.
New users benefit of a FREE trial of 300 vCPU/Hours to be used on the platform in order to test the platform, all each features and verify if it is suitable for their needs