Distributed Analysis

Distributed analysis allows users to split certain types of analyses and solve each portion of an analysis simultaneously on multiple machines. Simulation times can be greatly decreased by using this feature.

Note: When using distributed analysis, the Temp folder must be accessible by all machines set up in the HPC configuration.

Maxwell supports the following forms of distributed analysis:

Note: Communication between machines in remote analysis and distributed analysis can drastically affect performance. Use of a high-speed network system, like Gigabit or Infiniband, is recommended for optimal performance.

Maxwell Circuit and RMxprt designs do not support MPI. As a result, the supported forms of distribution depend on the design type.

All types of distribution for Maxwell 2D and Maxwell 3D use MPI, except when only distributing rows of a parametric table, either as a regular DSO or as Large Scale DSO performed through command line. MPI may also be used if Auto mode is specified, and rows of a parametric table are distributed. MPI must be correctly configured if the distributed analysis uses MPI. See Setting HPC and Analysis Options for setting MPI Licensing and the Ansys Electronics Desktop Installation Guide for details on installing MPI.

To configure a distributed analysis, you must select a distributed machine configuration containing a list of machines to use for a simulation, based on memory and CPU considerations (See: Selecting an Optimal Configuration for Distributed Analysis). To create an new distributed machine configuration or to edit an existing one, see Editing Distributed Machine Configurations.

Before you can select a configuration, it must be active. See: Setting HPC and Analysis Options.

To select an existing, active configuration:

Related Topics 

Configuring Distributed Analysis

Editing Distributed Machine Configurations

Selecting an Optimal Configuration for Distributed Analysis

Large Scale DSO for Parametric Analysis