In the Ansys Mechanical application, you can use the Resource Prediction option to predict the computing resources required to perform a solution. Predictions are based on hardware characteristics and factors such as model size. This functionality requires specific knowledge of the CPUs available on compute resources.
To support resource prediction when submitting solutions to Ansys HPC Platform Services, you can extract additional CPU details from the compute resources assigned to autoscaling cluster queues and make this information available to the Autoscaling Service and Resource Management Service.
The following CPU details are extracted:
CPU type (Intel/AMD)
CPU model (for example, 12th Gen Intel(R) Core(TM) i7-12850HX
Here is an overview of the process:
Start the Autoscaling Service with the --query-cluster (or -q) argument. For example:
sudo systemctl start hps_scaling_service.service --query-cluster
The Autoscaling Service creates a project in Ansys HPC Manager and submits a job to each queue specified in the "compute_resource_sets" section of the scaling_config.json file.
The account under which the Autoscaling Service submits the jobs depends on whether the Autoscaling Service is configured to run with impersonation.
If the Autoscaling Service runs with impersonation, jobs are submitted as the user mapped to the "service-account-rep-impersonation" user in the "user_mapping" section of the launcher_service.json file (see Configure the Process Launcher Service).
For example, if the following is specified in the launcher service file, the Autoscaling Service will submit the jobs as userA. Note that "service-account-rep-impersonation" should be mapped to a user who can query the job scheduler.
"user_mapping": {"service-account-rep-impersonation" : "userA"},If the Autoscaling Scaling service runs without impersonation, jobs are submitted using the credentials specified for the "username" and "password" properties in the scaling_config.json file (see Configure the Autoscaling Service).
Each job runs a shell script that collects CPU details using the
lscpucommand and writes the information to a file. The path of this file is specified in the "queue_info_path" property in the scaling_config.json file.
The Autoscaling Service monitors the queue_info_path file and updates the queue information with the additional CPU information.
When the Autoscaling Service registers with the Resource Management Service, the updated queue information is passed to the Resource Management Service.