Finite Element Analysis (FEA) is our core competency, our “bread and butter”. We use best-in-class commercial tools, open source offerings, and our own specialized solutions. Some clients approach us with a general problem in need of a solution and the scope of our engagement is the formulation of the problem, the solution, and the results. Other clients offload some of their computational requirements and we simply provide the computational infrastructure at very competitive rates.
FEA has always been (and probably will be) extremely compute bound. No matter the scale of the cluster, you will always have a problem that exceeds your current capacity. We strive to refresh our hardware every two years to maximize our computational capacity. Our compute clusters are more than adequate for the problems our clients bring to us. The performance of some applications scale almost linearly with the number of cpus (up to a point) if the appropriate network architecture is employed. Applications such as explicit dynamics, CFD, and iteratively solved diffusion (thermal) problems scale well in this environment. Implicit finite element solvers are more limited and generally perform best in a shared memory environment.
The larger the problem, the larger the dataset. Beyond the hardware plumbing required to deal with very large datasets, there is the data management problem. We have constructed a proprietary toolset that automates much of the handwork in assessing the results generated from these large models. Lightweight tools that provide a window on the data without requiring much resources themselves.