none
HPC Excel Workbook Offloading Slow Performance RRS feed

  • Question

  • Hi,

    I have been testing HPC Services for Excel as a possible way to speed up some slow-running Excel workbooks. However I have encountered a performance problem: Any cell-manipulating code inside the HPC_Execute method takes an unreasonable amount of time to execute when it is being run on a compute node through HPC (0.67 seconds), whereas if I run the workbook on that same node as if it were a regular workstation that time becomes negligible (0.00 seconds). The compute node is an Azure A2 VM running Windows Server 2012 R2 with Excel 2010 and HPC Pack 2012 R2.

    The workbook will (eventually) calculate the output of the workbook correctly, and the HPC Head Node is having no trouble splitting up the workbook into jobs for the nodes to process. I know that having cell-manipulating code in the HPC_Execute method is a bad idea, but removing that code will take too much effort and I can't find any reason as to why it would be running this slowly.

    Are there any known issues that would cause cell-manipulating VBA code to run slower than normal inside an Excel workbook on a HPC cluster?
    Wednesday, August 6, 2014 6:02 AM

All replies

  • Hi,

    There is no known issue for your situation.

    I suggest you contact Microsoft Support team and provider the Excel workbook to them to find the root cause.

    Thanks

    Friday, August 8, 2014 7:57 AM
  • Calling into Excel from the external code is slow.

    One approach we are exploring is to build all of your cell manipulation code into a VSTO Application level Add-In, and let it do the work.  It is *much* faster.

    Thursday, June 11, 2015 7:36 PM