This example demonstrates data flow tracking for an application graph with 2 root operators, 2 leaf operators, and 3 paths from the roots to the leaves. It also demonstrates the add_probe_operator function to track latency and message counts for intermediate operators.

The example showcases two key aspects of data flow tracking:

1. Application-level tracking: Using the DataFlowTracker API to monitor end-to-end latencies and message counts across the entire application graph.
2. Operator-level tracking: Using the get_data_flow_tracking_label() API within the PingMxOp operator to access message label information during compute, including:
   • Number of paths the message has traversed
   • Path names (comma-separated list of operators)
   • End-to-end latency for each path in milliseconds

Visit the SDK User Guide to learn more about the Data Flow Tracking feature.

• using deb package install or NGC container:

  /opt/nvidia/holoscan/examples/flow_tracker/cpp/flow_tracker

• source (dev container):

  ./run launch # optional: append `install` for install tree
  ./examples/flow_tracker/cpp/flow_tracker

• source (local env):

  ${BUILD_OR_INSTALL_DIR}/examples/flow_tracker/cpp/flow_tracker

• using python wheel:

  # [Prerequisite] Download example .py file below to `APP_DIR`
  # [Optional] Start the virtualenv where holoscan is installed
  python3 <APP_DIR>/flow_tracker.py

• from NGC container:

  python3 /opt/nvidia/holoscan/examples/flow_tracker/python/flow_tracker.py

• source (dev container):

  ./run launch # optional: append `install` for install tree
  python3 ./examples/flow_tracker/python/flow_tracker.py

• source (local env):

  export PYTHONPATH=${BUILD_OR_INSTALL_DIR}/python/lib
  python3 ${BUILD_OR_INSTALL_DIR}/examples/flow_tracker/python/flow_tracker.py