Composing simulated actors and running the simulation

The code below shows how to create and run a simulation-analysis in situ processing workflow using DTLMod to exchange data between the MPI_publisher and the subscriber that runs on a single core.

int main(int argc, char** argv) {
  sg4::Engine e(&argc, argv);
  e.load_platform("./platform.so");

  auto sub_host = e.get_host_by_name("node-0");
  auto pub_hosts = e.get_hosts_from_MPI_hostfile("./hostfile");

  // Create the data transport layer
  DTL::create("./DTL_config_file.json");

  // Start a simulated MPI code instance run by multiple actors
  SMPI_app_instance_start("MPI_publisher", MPI_publisher, pub_hosts, argc, argv);

  // Create a single in situ analysis actor
  e.add_actor("subscriber", sub_host, subscriber);

  // Run the simulation
  e.run();

  return 0;
}

This code first creates the SimGrid engine in charge of the execution of the simulation and loads a shared library that describes the simulated platform. Then, it assigns roles to hosts (i.e., simulated compute nodes) in that platform. The subscriber component is executed on node-0 while where the MPI ranks will run is defined in a regular MPI hostfile.

The DTL is created by calling DTL::create() (line~9). An instance of the MPI publisher and the subscriber actor are then created before running the simulation.

In situ processing workflows involving more than two components can be simulated using the same principles:

  1. each component is either an MPI code or a simulated actor;

  2. each component creates one or more streams to publish and/or subscribe to data;

  3. the streams create the flow dependencies between components and form the workflow.