Simulation Topology

Tags:

Overview

The Dynamic Simulation Environment provides several simulation topologies, all of which support distributed simulations, that may incorporate models running on multiple OS/Arch combinations or running in multiple simulation environments.

These topologies include:

  • Standalone - Stacked model instances running in a single process. Uses a Loopback SimBus.
  • Distributed - Distributed model instances, multi process, multi OS, optionally stacked. Uses a Redis SimBus.
  • Gateway - Remote simulation systems are connected with a gateway model. Uses a Redis SimBus.
  • MCL - Foreign models are imported via a Model Compatibility Library.
  • Realtime - Stacked model instances which may operate either Standalone or Distributed (i.e. Loopback or Redis based SimBus).
  • Embedded - External models and sensors and connected with a POP model / TCP

These topologies are supported by the Simer docker appliance (Linux) and ModelC runtime (Linux/Windows). The underlying messaging protocol used by the Dynamic Simulation Environment is implemented with Flatbuffers and may be implemented in any programming language or operating system.

Standalone

Several Models are stacked in a single model runtime. For example, the runtime may be packaged as an FMU using the FMI ModelC FMU wrapper.

Stacked w. Loopback

Characteristics:

  • Single Process - Sequential execution of models within a single process using the ModelC model runtime. Performance limited by CPU clock speed.
  • Co-Simulation - Phase shift for signal exchange between model steps (i.e. output to input). Phase shift is equal to one simulation step.
  • Loopback SimBus - Low latency execution with no external/network signal exchange.
  • Limited deployment - Use of Loopback SimBus means that only single process deployment is possible.

Note: Models may be stacked in any deployment topology (i.e. with Redis).

Sequence Diagram

topology-standalone-stacked

Configuration

CLI / Environment
VariableCLI Option (ModelC/Simer)Value
SIMBUS_TRANSPORT--transportloopback
SIMBUS_URI--uriloopback
Simulation Stack
simulation.yaml
---
kind: Stack
spec:
  connection:
    transport:
      loopback:
        uri: loopback
  runtime:
    stacked: true

Stacked Sequential Co-Sim

Characteristics:

  • Single Process - Sequential execution of models within a single process using the ModelC model runtime. Performance limited by CPU clock speed.
  • Co-Simulation - Sequential Co-Simulation resulting in no phase shift of scalar signals exchanged between models in the stack.
  • Limited deployment - Used to compose models from several “component” models where the resultant model does not exhibit phase shift between its input and output scalar signals for any simulation step.

Sequence Diagram

topology-standalone-sequential

Configuration

CLI / Environment
VariableCLI Option (ModelC/Simer)Value
SIMBUS_TRANSPORT--transportloopback
SIMBUS_URI--uriloopback
Simulation Stack
simulation.yaml
---
kind: Stack
spec:
  connection:
    transport:
      loopback:
        uri: loopback
  runtime:
    stacked: true
    sequential: true

Distributed

Models are distributed in several instances of a model runtime. Model runtimes may run on several computers, use different operating systems, and execute in other Simulation Environments. Models can also be stacked, and those stacks can be configured to operate in Sequential Co-Sim mode.

Distributed w. Redis

Characteristics:

  • Multi Process - Parallel execution of models using a Simer/ModelC model runtime. Performance limited by number of CPU Cores and Processing Nodes (e.g. VMs/Workstations).
  • Co-Simulation - Phase shift for signal exchange between model steps (i.e. output to input). Phase shift is equal to one simulation step.
  • Redis w. SimBus - Low latency signal exchange when running with a Simer deployment, medium latency when operating cross-platform simulation (latency depends on network performance).
  • Flexible deployment - The Simer Container Runtime can be used for high performance, scaled, Linux deployments. The underlying model runtime ModelC is available for Windows deployments. Both can be combined to create distributed simulations.

Sequence Diagram

topology-distributed-redis

Configuration

CLI / Environment
VariableCLI Option (ModelC/Simer)Value
SIMBUS_TRANSPORT--transportredis
SIMBUS_URI--uriredis://localhost:6379

Note: There are several variations of Redis transports, redispubsub being the most common (and original). The newer redis transport has the best performance and is suggested for new simulation projects.

Simulation Stack
simulation.yaml
---
kind: Stack
spec:
  connection:
    transport:
      redis:
        uri: redis://localhost:6379

Gateway

Remote Simulations are connected to a DSE/Simer based Simulation using a Gateway Model. The Gateway Model is implemented in the Remote Simulations framework/environment which then connects to the DSE/Simer Simulation. The connection may be established with either an existing integration (i.e. the DSE ModelC Library), or a direct implementation of the underlying DSE messaging protocols.

ModelC based Gateway

Sequence Diagram

topology-gateway-modelc

MCL

Foreign models are imported to a simulation using a Model Compatibility Library which may be implemented for any model standard (e.g. FMI MCL).

Realtime

A simulation is deployed in a realtime simulation environment (e.g. HIL) with a combination of distributed and stacked model instances, using a Redis based SimBus, to ensure consistent realtime operation.

Embedded

External models, sensors and ECUs are connected via native protocols (e.g. TCP) to a Point-of-Presence model running within a Simer based simulation.