Running on a Linux Cluster Using Pre-Allocation

In pre-allocation mode, Design Manager allocates all necessary resources and licenses before the compute simulations begin. You submit the whole Design Manager project to the cluster queue in one step. The pre-allocation mode is supported only for Linux clusters.

To submit the entire Design Manager project as a single job to the queue of your cluster:

  1. A pre-allocation mode requires a Direct compute resource. For each design study in your Design Manager project, set the run and resource settings:
    1. To specify the default direct compute resource, right-click the Tools > Compute Resources > Default Direct node and set the properties described in Compute Resource Reference—Direct.
      You could also create an additional user-defined Direct compute resource to assign to Compute Resource for Design Run within the Run Settings > Single Resource node later on.
    2. Select the Design Studies > [design study] > Settings > Run Settings node and set Resource Mode to Single Resource and specify the following properties:
      When setting the numbers for Simultaneous Jobs and Compute Processes, check that you have enough cores available (simultaneous jobs * compute processes).
    3. Select the Run Settings > Single Resource node and select the Default Direct or another user-defined [Direct] compute resource as Compute Resource for Design Run
  2. Select File > Save and save the project file.
  3. Transfer the Design Manager project file (*.dmprj) and associated input files such as the reference simulation file (*.sim) and custom java macro files (*.java) to your cluster project root directory.
  4. Prepare your cluster job submission script to meet the Pre-Allocation Submission Script Requirements. Make sure that the script file has execute permission.
  5. From your cluster project root directory, submit the Design Manager project using the appropriate queue submission command for the cluster.
    Example: qsub -l nodes=6:ppn=16 <submissionScript>.sh, where -l nodes=6:ppn=16 specifies the required resources for the simultaneous design simulations.
The job scheduling system on your cluster allocates the resources, such as 6 * 16 cores = 96 cores in the example above. Within the allocated resources, the Design Manager server starts on the first assigned compute node and acquires the resources to submit the design simulations one by one. Each design simulation requests the number of cores that you specify in the Run Settings.

The Design Manager server ensures that you do not overload the allocated resources. Suppose you specify 4 simultaneous design simulations and 32 compute processes for each. The Design Manager server runs the first 3 design simulations using 96 cores (3*32) and starts the 4th simulation as soon as one of the first three completes.

When submitting the Design Manager project, make sure that you do not allocated resources inefficiently. Suppose you specify 2 simultaneous design simulations and 32 compute processes to run with each design simulation, then 32 of the 96 allocated cores remain unused until the design study completes.

  1. To follow the progress of your design studies, you can connect to and disconnect from the running Design Manager server, see Making Client-Server Connections.