DPF Crashes Workbench with large result files
I've noticed an issue when using DPF with large result files in my ACT scripts (currently using version 2023R2), essentially whenever the result file gets large enough (25+GB) all of my post processing scripts crash workbench a few seconds after starting to run. I reduced the size of mesh and resolved to get a smaller results file and had no issues.
Has anyone else noticed this issue / have any solutions for how to deal with it? It seems like it's a memory problem, however I am using a "reasonable" machine for our organization (20 Core, 96GB RAM, Solid State Drive). So essentially adding RAM to every users machine is not a viable solution.
Comments
-
This seems like something to report in the ANSYS official support channel as it relates to a crash of an application.
That said, if you are trying to post process very large data sets, memory can become an issue. Practically, as you mentioned, it isn’t feasible to give every user infinite memory, so you should start to implement memory monitoring aspects of your scripts. You can reduce the memory DPF is using by reducing the mesh scoping domain or the time/freq domain. For example maybe only post half the time, then the other half, or half the parts, etc….
If you try to implement the above and still see crashing or memory issues, that again is a good candidate for an official support request.
0 -
I did some more investigating and tried chunking the data and processing it separately. The issue seems to be less with the size of the results file, and more with the size of the fields container. As a fields container grows in size it eventually hits a point where it crashes workbench. A model can actually be relatively small in size, but if there are many time steps eventually the fields container cannot handle the data.
0 -
@Nick Ramseyer This was indeed an issue fixed in 2024R1 DPF Server. You can wait for the release of 2024R1 which should happen at some point in this month or use the
2024 R1-pre0
standalone DPF server:0 -
Hey Ayush! Thanks! I'll give it a go when the new version releases.
1