Running a simulation with Distributed Rendering

After configuring Distributed Rendering and tagging the GameObjects in the scene to suit our needs, we are now ready to run the simulation!

To launch a simulation with Distributed Rendering, you’ll first need to create a Player Build. Then you’ll need to launch multiple instances of this build, where each instance is launched as either the Server (maximum of one instance) or as a Client.

Launching the instances

  1. Build the application

    • File -> Build Settings -> Select Target Platform -> Build

  2. Open the Terminal or Command Line at the location of the application

  3. Start the application with no arguments to run as the Server.

    • Add the options -batchmode -nographics on Linux and Windows to run the Server without any graphical or window system overhead

      $ <build-name.x86_64> -logFile <log-path>

  4. Start the application with the -render argument to run as the Client

    • Add the option -ip <address> with the IP address of the Server for <address> to establish an explicit connection 

      $ <build-name>.x86_64 -render -group-id 1 -logFile <log-path> & ... & <build-name>.x86_64 -render -group-id N -logFile <log-path>

      If using VMs, include the --ip <server's internal IP address> parameter.

Warehouse Robotics scene run with Distributed Rendering

💡 Note: In case Camera Affinity Mode or Scene Serialization Mode is turned on, start as many Clients as the number of camera nodes in the scene. However, in case Lock Step Sync Mode is activated, start as many Clients as one less than the number of camera nodes in the scene (because the Server renders one of the camera nodes).

The Client Launcher

As the number of Clients and configurations increase, keeping track of Clients launched can get tricky. To solve this, the Distributed Rendering package comes with a Client Launcher module that makes the process of running a Server and Clients a lot smoother. Using this component, Clients can be directly launched from the Editor or from a build stored on the system without specifying any arguments using Terminal or Command Line.

💡 Note: The ClientLauncher component is not currently supported in IL2CPP builds.

Warehouse Robotics scene run with Distributed Rendering

This component can be added to any GameObject in the scene. It contains the following options:

  • Run: Launch the Server and Client(s). Not accessible if Launch On Play is selected.

  • Auto Configure: This fills in the launcher module with information that most suits the project based on the current settings. This includes filling in the Client Executable location and the Number Of Clients based on the number of cameras tagged in the scene and the Server mode. Make sure to hit this button after any major changes in the scene that may affect Distributed Rendering.

  • Browse for client executable...: This lets the user browse their system for an executable file for this project, which is then set in the Client Executable field.

  • Select last build: This automatically sets the Client Executable field with the location of the last built executable.

  • Find Subscribers: Displays the number and names of cameras tagged with a Camera Mode component.

  • Kill _ running client processes: Kills any running Client processes.

  • Launch On Play: Toggle to launch Client(s) when Play mode is activated. The -launch-clients command line option can be used to override this externally.

  • Client Executable: Location of an executable of the current project.

  • Additional Arguments: Any additional arguments the user may want to specify.

  • Hide Windows: Runs the Clients with the -batchMode switch, so that they don't display a window.

  • Number Of Clients: Number of Clients to run.

  • Capture Client Logs: Sends the Client logs back to the Server and to be displayed in the Editor Console.

  • Per Node Parameters: Provides a list of unique parameters to the command line for each Client invocation.

Command line options

While executing the simulation, the following command line options are also available for instance specific configurations:

  • --help

    Log DR help text, then exit.

  • --distributed-rendering, --no-distributed-rendering

    Run the simulation with Distributed Rendering.

  • --render, --no-render

    Identify the running instance as a Client. Any instance that is not the unique Server should include this flag when launched.

  • --pause-at-startup, --no-pause-at-startup

    Pause the simulation at startup.

  • --physics, --no-physics

    Enable physics system update.

  • --frame-batching, --no-frame-batching

    Enable sending multiple frames to a Client at a time.

  • --frame-batch-load-balancing, --no-frame-batch-load-balancing

    Enable load balanching for frame batching (only when running in Asynchronous mode).

  • --frame-batch-size <integer>

    Maximum number of frames to send in a single batch (only when frame batching is enabled).

  • --debug-ui, --no-debug-ui

    Enable the debug UI display.

  • --server-address <address (string)>

    For Clients, indicates the IP Address or host name of the Server.

  • --ip <server's internal IP address>

    For Clients, creates an explicit connection between the Client and the Server. This is especially useful when using VMs.

  • --wait-for-clients, --no-wait-for-clients

    Override the server WaitForClients setting.

  • --frame-count <integer>

    Maximum number of frames to render before exiting (set to 0 to run endlessly). While this is useful for debugging, it should be avoided when running in an environment without shell access.

  • --network-port <UInt16>

    Overrides the "Network Port" option. Network address port where the socket connection will be established between the Server and Clients.

  • --discovery-ping-interval <float>

    Override the interval in seconds that the UDP cluster will ping for auto-discovery of other Clients on the local network. This is unused in Clients when connecting directly to a Server at a specific address using the --ip option.

  • --group-id <integer>

    User-defined group ID to assign to the running Client instance (ignored on Server instance).

  • --workload-id <integer>

    User-defined workload ID to assign to the running Client instance (ignored on Server instance). Any value that is 0 or less is considered "not set" and not providing this option when launching a Client is the equivalent of passing --workload-id 0.

  • --client-name <string>

    User-defined name to assign to the running Client instance (ignored on Server instance).

  • --render-node-count <integer>

    User-defined number of expected render nodes to be launched while running in lock-step mode (defaults to number of tracked cameras).

  • --connection-timeout <integer>

    Client connection to server timeout, in milliseconds. Client instance quits if timer expires and no connection has been made (ignored on scene server instances). Zero indicates no timeout.

  • --launch-clients, --no-launch-clients

    If ClientLauncher component is active in the Scene Server, override its Launch On Play setting.

  • --sync-mode <mode (string)>

    Sets the Server sync mode to one of four possible options: asynchronous, camera-affinity, lock-step and scene-serialization.