Hi @witor,
you are right, the error message and its recommendation is not very clear. Let my try to explain it.
Every time you start a simulation run on SimScale a cloud computing instance with the specifications as defined under the “Number of computing cores” is fired up. It defines how many cores the instance has and implicitly also how much RAM is available, according to this table: https://www.simscale.com/docs/content/simulation/simulationControl/numberOfCPUs.html
For some solvers available in solid mechanics analysis types at SimScale (MUMPS and PETSC) the parallelization of the computation is done in a way that the initial computing domain (mesh) is divided into smaller chunks and on each one the solution is computed “independently” (not on the boundaries). This means the total available RAM is devided by the number of these parallel runs. This number is defined under “Number of cores used for the computation” (the wording is indeed misleading and we probably will change it). So, if your analysis runs out of Memory (RAM), there are essentially two options:
- increase the available RAM on the machine (by increasing the "Number of computing cores) and leaving the “Number of cores used for the simulation” at the same level, or
- leave the machine size at the same level and reduce the “Number of cores used for the Simulation”.
Of course this is only a simplification, as subdividing the computation domain into more partitions will also reduce the memory requirements to solve each domain as the get smaller, but this does not scale linearly.
Coming back to your problem. The above actually did not fully apply to your situation as you were using the MultFront solver, which does not use this kind of parallelization and is generally less memory efficient than MUMPS.
I made a copy of your problem and changed the solver to MUMPS and used a 32 core instance and it works just fine.
you can check out the project here: https://www.simscale.com/projects/rszoeke/printer_x_axis/
Best,
Richard