In this article, we will explain how to launch parametric sweeps/optimizations on a (1) local machine, (2) across multiple computers in your local network, and (3) with a job scheduler on a cluster. Running these types of tasks concurrently can drastically shorten the time-to-result for optimization/sweeps with multiple jobs.
Note:
- Machines/nodes should be on the same operating system and using the same version of our simulation design software.
- Configure your system first, as shown in the Configuring your cluster for Ansys Lumerical page.
- Running simulation jobs at the same time (concurrent) on multiple machines will require an additional license for each computer that is running the job.
- Before trying to run your simulation or sweeps across several computers, ensure that you can run this on each computer independently.
Setting the Resource Configuration
- Open the Resource Configuration from the toolbar or Simulation menu.
- Enter the number of concurrent sweeps/analyses you want to run on your machine in the Capacity field. This dictates the total number of available resources of that type when running sweeps/analyses.
- You can also use the Duplicate button to create multiple resources. For example, duplicating a resource with a capacity of 2 will give you a total of 4 resources. Duplicate it again and you have 6 total resources.
- Configure the Processes and Threads settings applied to each of the engines. Note that not all solvers support Processes or Threads. For those that support both, finding the right combination often comes down to your specific simulation needs.
- See the Resource Configuration page for details.
Default job launching preset
Windows:
- Use "Remote: Intel MPI" as the job launching preset. See this article for details.
- Select "Custom" for custom MPI installations. e.g. clusters or HPC that have a preconfigured MPI implementation.
- Do Not Use "Remote: Microsoft MPI" or "Local Computer" as the job launching preset.
Linux:
- Use "Remote: MPICH2" to test with the include MPICH2 runtime. Note: this runtime is outdated and is not as performant as Intel MPI or OpenMPI (see "Custom" bellow).
- Select "Custom" if for custom MPI installations e.g. clusters or HPC that have a preconfigured MPI implementation.
- Do Not Use "Local Computer" as the job launching preset.
Calculating how many licenses are required
The number of FDTD accelerator, solve, or HPC licenses required to run your jobs depends on the number of cores used, the number of concurrent simulations, the type of job, and the type of license you have purchased. See the Understanding solver, accelerator, and HPC license consumption page for details.
Resource configuration examples
Run in parallel on a single machine
Run concurrently across multiple machines
See also
Lumerical job scheduler integration configuration (Slurm, Torque, LSF, SGE)