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_regressionDocker 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_classificatiobenchmark in a Docker container and collects GPU metrics at a 10-second interval. The--carbon_regionflag 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_logsflag 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 usingtmuxinstead 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 intmuxallows 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