Our transition to Cloud Computing means we can run CFD simulations on greater computing power with effectively endless capacity. This means we can run more detailed models, with quicker computing speed and can even solve several simulations in parallel. This accelerated turnaround of high detail data allows more informed design decisions with no delay to projects.
And guess what… all these benefits come with no extra costs.
Engineering simulations have evolved greatly over the last 2 decades.
Once upon a time, not that long ago, creating a mesh for a CFD simulation was a relatively manual process. A fine balance of creating a quality mesh avoiding bad cells, that was fine enough for simulation granularity yet still able to run on the available hardware. Spending weeks generating a mesh was not untypical, tweaking individual node locations until the mesh satisfied the necessary criteria. Set your model solving and come back next week to view your results.
Hardware and meshing software colluded to make the CFD engineer the most frustrated person in the office.
This scenario is now unrecognisable.
Hardware, once the biggest bottleneck in engineering simulation, is now only a minor consideration when setting up simulations. This is especially true now that Cloud Computing is so widely accessible and affordable.
At ECS, we have been running our trusty internal High-Performance Computer for several years, churning away simulations. This was a significant improvement on the workstation solutions we employed prior to that. The increase in memory (the most significant bottleneck for CFD model size and detail) from 128GB to 576GB allowed greater detail to be captured in our simulations, providing greater insight into designs and opening opportunities for better optimisation of flow fields and operating temperatures.
The benefits of greater computing power extend beyond the allowable simulation model size and speed. Our automated mesh processes significantly reduced the user input necessary to set up and solve simulations considerably.
Yet, we found that we were still occasionally exceeding the memory limits of the cluster, crashing simulations. Moreover, jobs were being queued up as we had jobs queuing up to run on our single HPC.
This year, we have gone full Cloud!
Operating through AWS, we have access to a wide range of state-of-the-art computing capacity. Now we can run a simulation across several hundred processors with several TB of RAM.
Those of you with commercial CFD licenses will need to pay big bucks to unlock additional computing power. Our proprietary CFD process built on extensively validated open-source solvers allow us to access all this computing power with no additional license costs.
Jealous? You should be.
The convenient Pay as You Go leasing costs of this computing power means we pay only when we are actively using it. Optimising the process to use the right amount of computing power for each simulation keeps costs down. Compared to our HPC running costs (including electricity and maintenance), cost per simulation is almost identical.
Our customers experience many benefits from this transition to Cloud Computing:
- Greater detail: more extensive analysis of simulation results with greater detail
- More Informed: multiple simulations simultaneously for comparison
- Better decisions: greater confidence from more data
- Quicker results: turnaround of simulations is quicker
Get in touch to hear how you can benefit from our highly optimised simulation capability for time and cost effective thermal design improvements.