Skip to content

Example Commands

Example Usage

Example 1: Basic Monitoring with Completion Plot:

iris-gpubench --benchmark_image "synthetic_regression"
  • Explanation: This command runs GPU monitoring while executing the benchmark specified by the Docker image synthetic_regression. The system will collect GPU metrics and generate a completion plot at the end. Live monitoring of GPU metrics is enabled by default.

Example 2: Exporting Data to VictoriaMetrics:

iris-gpubench --benchmark_image "synthetic_regression" --export_to_meerkat
  • Explanation: Similar to the first example, this command runs the synthetic_regression Docker image benchmark and collects GPU metrics. Additionally, the collected data is exported to Meerkat for long-term storage and further analysis. This is useful when you need to monitor metrics over time and visualize them later using external tools such as the Grafana Dashboard.

Example 3: Full Command with All Options:

iris-gpubench --benchmark_image "stemdl_classification" --interval 10 --carbon_region "South England" --live_plot --export_to_meerkat --monitor_logs
  • Explanation: This is a comprehensive example that runs the stemdl_classificatio benchmark in a Docker container and collects GPU metrics at a 10-second interval. The --carbon_region flag specifies the carbon intensity region as "South England" to track the carbon emissions impact. Live plotting of GPU metrics is enabled (--live_plot), and data will be exported to Meerkat DB via VictoriaMetrics (--export_to_meerkat). The --monitor_logs flag enables logging of both GPU metrics and the Docker container logs, allowing for deeper analysis of benchmark performance.

Example 4: Run and Monitor Benchmark in the Background without the Need for a Container:

/mantid_imaging_cloud_bench$ iris-gpubench --benchmark_command "./run_1.sh" --live_plot --interval 1
  • Explanation: In this example, a benchmark command (./run_1.sh) is executed in the background using tmux instead of a Docker container. GPU metrics are collected at 1-second intervals, and live plotting of these metrics is enabled. This is useful when you have a script or binary that doesn't require containerization and want to monitor the system's GPU usage in real-time. Running benchmarks in tmux allows the process to continue in the background, making it ideal for long-running benchmarks that don't need constant attention.
  • Important: For this example, you'll need to install you benchmark on the VM and the iris-gpubench package.

Previous Page | Index | Next Page