Distribution of Computing Resources Across System Coupling Participants

In a project with System Coupling, the individual components (such as Fluent, Mechanical, or CFX) share resources. By default, when design points from a project containing a System Coupling system are submitted to RSM, available computing resources are distributed equally across all coupling participants according to the number of compute nodes available and the number of processes requested. Component-level parallel settings (such as the Execution Mode and Number of Processes properties) are disregarded during the design point update.

Example 1: Six Processes

If a coupled simulation requests six processes to be executed on a cluster that has six nodes with one processor per node, then participant solvers will start six processes, with one on each compute node. The participant solvers alternate their processing, so that only one participant is actively using computing resources at a time.

NodeSolver Processes for Participant 1 Solver Processes for Participant 2

1

1

1

2

1

1

3

1

1

4

1

1

5

1

1

6

1

1


Example 2: Ten Processes

If a coupled simulation requests 10 processes to be executed on a cluster that has 4 nodes with 8 processors per node, then each participant solver will start 10 processes.

NodeSolver Processes for Participant 1 Solver Processes for Participant 2

1

8

8

2

2

2

3

0

0

4

0

0