Platform

Platform

Topics related to optiSLang, HPC, DesignXplorer, Cloud and more.

Ansys Distributed HPC: Memory allocation

    • anarayanan
      Subscriber

      Dear Community,

      I am solving a modal analysis problem on a linux cluster (version 19.1). I have 128 GB memory on the node distributed on 20 cores. I have allotted 16 cores for the solution and am expecting a minimum of ~90 GB memory available for it. However, in the solution output file reads Maximum total memory allotted is ~ 30GB and Total physical memory available is 126GB.

      When I include (-m value (mb)) command in the Ansys command line I get Error stating: The memory (-m) size requested [value mb] is not currently available.  Reenter ANSYS command line with less memory requested.

      My eigen solver requires more than 32GB memory to converge to a solution. I have attached distributed ansys statistics to this thread.

    • George Karnos
      Ansys Employee
      Is the specific issue that you can not allocate more than 32GB or memory? Are there OS user limits set on the linux machine?
      Can you paste the whole out file? Please paste the text as staff are not permitted to open files.
    • anarayanan
      Subscriber
      Dear Geo I have rectified the issue. It seemed that Ansys was not requesting memory (that was available on the Linux machine). However, when I changed the solution method to Direct instead of Iterative it seemed to have solved this problem.
      Regards Ashvij
    • mrife
      Ansys Employee
      Hi @anarayanan Remember that the iterative solvers, like PCG Lanczos, require about an order of magnitude less memory than the direct solver does. Also when manually allocating the memory on the command line the value is per core, and not total, when using distributed parallel.
      Mike

Viewing 3 reply threads
  • The topic ‘Ansys Distributed HPC: Memory allocation’ is closed to new replies.