Run a Case Study Simulation

The supply of computing resources changes based on demand, and customers want the most cost effective usage of those computing resources. Simulations or case studies provide key metrics such as core utilization and job throughput for determining the most productive way configure an HPC.

Before you can begin running simulations, you must first create an HPC snapshot. Once the snapshot is imported, it can then be used for successive simulations. Some degree of validation is done when importing a snapshot and snapshots can be rejected if the snapshot contains options, attributes, or features that make the simulation behavior unpredictable. Review the following information for a list of reasons for a snapshot rejection.

Tip: Use the navigation links on the left hand side of the web page to quickly jump to a section of the New Simulation form.

Create a Case Study Simulation

Start a simulation by providing a name and a description.

  1. Click the Simulate tab.
  2. Click New Simulation.
  3. Click located in the upper right-hand corner of the web page.
  4. Enable the Display Advanced Settings check box.
    If this setting is disabled, then a subset of Scheduler parameters is displayed for the HPC snapshot.
  5. For Type, enable Case Study.
  6. For Label, enter a name for the case study.

    The name can be comprised of numbers, letters, and underscores, however at least one alphanumeric character is required.

  7. For Description, enter a description for the case study.

Import or Select an HPC Snapshot

Import a previously created HPC snapshot or select a snapshot that has previously been imported.

Before you can begin running simulations, you must first create an HPC snapshot. Once the snapshot is imported, it can then be used for successive simulations. Some degree of validation is done when importing a snapshot and snapshots can be rejected if the snapshot contains options, attributes, or features that make the simulation behavior unpredictable. Review the following information for a list of reasons for a snapshot rejection.

  1. Optional: To run a simulation for an HPC snapshot that has not yet been imported:
    1. Click or the Import a Snapshot link.

    Add a Snapshot
    Figure 1. Add a Snapshot
    1. For Snapshot Name, enter a name for the snapshot.
      The snapshot name can be comprised of numbers, letters, hyphens, and underscores.
    2. For Description, enter a description of the snapshot.
    3. For Snapshot, select the tarfile that represents a snapshot of your HPC cluster.
    4. Click Submit.
      Depending on the size of the tarfile, it may take some time to process the snapshot.
      Once the processing is complete, Workload dates, Node Classes, queues, limits and PBS Professional Scheduling parameters associated with the snapshot are populated.
  2. Otherwise, to run a simulation for an HPC snapshot that has already been imported, select the HPC snapshot from the list.

    Choose HPC snapshot
    Figure 2. Choose HPC Snapshot
    Once the snapshot is selected, Workload dates, Node Classes, queues, limits, and PBS ProfessionalScheduling parameters associated with the snapshot are populated.

Choose a Workload Interval

Select a specific workload time interval. Jobs that ran during that time interval are included in the simulation.

The default workload interval for the simulation is based on the job's included in the HPC snapshot. If the snapshot includes jobs that ran between January 1st, 2019 and March 30th, 2019, then the default workload interval is set to that time period. The simulation can be run to include the entire snapshot workload interval (i.e., all jobs) or can include a narrower time period (i.e., only those jobs that ran during the specified time).
  1. Click Workload from the PBS Professional menu located on the left-hand side of the web page.
  2. Choose a workload interval by selecting a start and end date:
    1. Select a start date by clicking to the right of Start Date.
    2. Choose a start date.
    3. Select an end date by clicking to the right of End Date.
    4. Choose an end date.

Change the HPC's Node Configuration

Change the HPC's system configuration by increasing or decreasing a grouping of execution nodes called Node Classes.

Execution nodes are grouped based on RAM, CPU and GPU signatures, that is execution nodes having the same number of CPUs, GPUs and RAM are grouped into a Node Class.
  1. Click Node Classes from the PBS Professional menu located on the left-hand side of the web page.
  2. Scroll through the list until you find the grouping that you want to either scale up or down.
    Figure 3. Available Node Classes
  3. Click to the right of the Node Class that you want to alter.
  4. Change any of the following values:
    Value Action
    Label Change the name of the Node Class.
    NCPUS Increase or decrease the number of CPUs.
    Memory (units: B, KB, MB, GB, TB, PB) Increase or decrease the amount of RAM.
    GPUS Increase or decrease the number of GPUs.
    Count Increase or decrease the number of execution hosts.
  5. Click to save the changes.
  6. Repeat step 2 through 5 to alter additional Node Classes.

Change the Queue Configuration

Modify the HPC cluster's execution queue configuration.

Only execution queues can be modified. Snapshots containing routing queues are rejected. For more information see Snapshot Rejection.
  1. Click Queues from the PBS Professional menu located on the left-hand side of the web page.
  2. Click located to the right of the queue you want to change.

    Queues
    Figure 4. Queues
  3. For Priority, set the queue's priority.
    The valid range for a queue's priority is -1024 to 1023. A higher value for priority means the queue has greater priority.
  4. For En., choose one of the following options:
    • Enable the En. check box to allow the queue to accept jobs, meaning that jobs can be enqueued in the queue.
    • Disable the En. check box to disallow the queue from accepting jobs.
  5. For St., choose one of the following options:
    • Enable the St. check box to start the queue, allowing the jobs in the queue to be executed.
    • Disable the St. check box to stop the queue, disallowing jobs in the queue from running.
  6. Click to save the changes.

Change the Resource Usage Limits

Add, edit or delete resource usage limits for the simulation.

  1. Click Limits from the PBS Professional menu located on the left-hand side of the web page.
  2. Change the resource usage limits for the simulation by adding, editing or removing resource usage limits.

Change the Workload Manager Scheduling Parameters

Scheduling parameters can be modified for a simulation.

  1. Click Configuration from the PBS Professional menu located on the left-hand side of the web page.
  2. Change any of the scheduling parameters:
    1. Click located to the right of the parameter.
    2. Change the value of the parameter.
    3. Click to save the changes.

Start the Simulation

Submit the simulation. Once the simulation is complete you can view the simulation results.

  1. Click Start Simulation.
    A message is displayed stating that the simulation is starting and a bar is displayed showing the simulation progression.
    Tip: If you do not want to wait for the simulation to complete, you can navigate away from the New Simulation page and view the results of the simulation at a later time.
    Once the simulation is complete, the simulation results are displayed at the top of the web page.
  2. Choose one of the following options:
    • View the simulation analysis charts.
    • Click the View the full analysis link to view the charts via the Analyze tab.
    • Scroll to the bottom of the web page and click New Simulation to run a new simulation.