Simr Blog

How to select a Cloud Provider for CAE Simulations

Written by David Waltzman | Jan 24, 2019 9:04:32 PM
Your simulations have exceeded the capabilities of your desktop workstation. Now what?! Project workloads are only intensifying, requiring more computing resources than the IT department is willing to buy for your group. If only there was a cloud service for running simulations…

 

Running computer aided engineering in the cloud is not brand new in 2019. However, running cloud FEA is becoming more and more feasible for engineers from both a cost and security perspective. The paradigm shift of cloud simulation is that now the IT resources are sized to fit the engineering questions you want to solve, rather than the other way around.

 

Simulations are a subset of the broader category of high performance computing. Usual suspects also included in the HPC bucket are big data analytics, AI/ML/DL, and rendering. Cloud HPC significantly lowers the barrier to entry for these advanced fields that otherwise would have required significant upfront investment in servers. HPC on demand means that now you only need to pay for world-class purpose-built compute clusters when you need it. No more collecting dust on the shelf. It also means that the hardware spec is not defined by one project then manipulated to fit the bill for future projects.

 

All of that being said, how to do I choose the right cloud for CAE?

 

It depends :)

 

Most large companies are advancing in their digital transformation journey to where they have an enterprise cloud strategy. My advice would be to ask to your company IT if you are at that level of maturity. If yes, then I would ask if engineering has access to it. If no, then I would ask to be a part of the selection process so that the chosen cloud provider is a good fit.

 

My personal bias is towards Microsoft Azure. They were the first to really focus on optimized hardware for FEA and CFD with high-end GPUs (accelerated graphics and computation) and Infiniband (fast communication between nodes ideal for scaling MPI jobs). Their focus on the manufacturing space with IoT also makes them a great candidate as Digital Twins become more commonplace.

 

One thing is for certain. Cloud offerings are constantly evolving. Knowing the differences between Azure, AWS, and GCP today does not mean the same will hold true a year from now. My suggestion is to do your homework and stay up to date to monitor which cloud is most aligned with your business.