Moldflow insight Linux HPC Inquiry

[Jae_Hoon_Lee]

Hello, Moldflow users.

I do not know Linux. But you have to use HPC.

So please check some things.

1. Can I use only DOE or Optimization?
Is public analysis not available at all?

2. If public analysis is available,
can one study be used with multiple nodes?

3. Must use the pbs file to use HPC,
Please tell me how to use it.
Ex)
HPC has a total of eight nodes.
enode0 (Hade), enode1 ~ enode7
enode0 = 4 CPUs 24Core
enode1 ~ 5 = 4 CPUs 24Core
enode6 ~ 7 = 2 CPUs 12Core
Test.sdy-> (No DOE, No Optimization)
Analysis Node-> enode5

Please tell me how to create the pbs file is these cases.

[juan.zamora.ruiz · Accepted solution]

Hi @Jae_Hoon_Lee ,

Thank you for participating in the support forums. I will try to answer your questions and provide some additional resources for information.

1. Can I use only DOE or Optimization? Is public analysis not available at all?

No, you can run any type of analysis on a HPC setup. DoE, Optimization and runner balance are best suited for HPC, but you can run any type of analysis.

2. If public analysis is available, can one study be used with multiple nodes?

Assuming that public analysis stands for the rest of analyses, not DoE, Optimization or runner balance, the answer is no. The solver will run on a single node. The solvers are threaded, not Message Passing Interface (MPI) enabled, and use as many threads as are physically present on the execution machine, by default. For jobs that spawn child jobs, such as DoE, Optimization, or Runner Balance Studies, the SJM can distribute these child studies to other nodes, in order to spread the workload.

3. Must use the pbs file to use HPC, Please tell me how to use it.

There's an example.pbs script deployed with the Liniux Moldflow installation. You will find it within the installation folders subfolder autodesk-moldflow-ami2019/bin. You will edit it and adapt the variables in the PBS section to match your HPC setup. As this is system specific, the system administrator for your HPC system will be best placed to inform you the precise syntax required.

You can read the help page on High Performance Clusters (HPC).

Regards