Transport protocol

This criterion defines the chosen protocol to effectively implement on target computers the communication between models. The choice of a transport protocol is independent of the type of simulation or the step size.

Direct link

This is the simplest implementation: the two models are linked together (using for example a shared library) in the same executable and executed within a unique process on the same computer.

Example: a Simcenter Amesim model is compiled as an S-function directly linked with Simulink.

Local interprocess communication

Each model is executed within a distinct process, on the same computer. Processes communicate by means of a local protocol, like pipes or shared memories.

Example: the discrete partitioning interface. A Simcenter Amesim master model communicates with a Simcenter Amesim slave model on the same computer through pipes.

Network protocols

Each model is executed within a distinct process, on one or many computers. Processes communicate by means of a network protocol. This protocol is generally based upon IP (this may be a proprietary protocol over TCP or UDP, or XML, SOAP, etc.), but may be a field bus protocol like CAN.

Example: a Simcenter Amesim model communicates with another Simcenter Amesim model by means of submodels implementing shared memories.

Message passing protocols

This is a generalization of local inter-process and network protocols. Each model is executed at a node of a logical cluster of processing resources relying on message passing protocols like PVM, MPI or OpenMP to implement local or network communications if needed.

Example: same example as above, with submodels implementing MPI.