Monitoraggio delle prestazioni della GPU sulle VM Windows
Mantieni tutto organizzato con le raccolte
Salva e classifica i contenuti in base alle tue preferenze.
Windows
Per contribuire a un migliore utilizzo delle risorse, puoi monitorare il tasso di utilizzo della GPU delle tue istanze di macchine virtuali (VM).
Quando conosci i tassi di utilizzo delle GPU,
puoi eseguire attività come
la configurazione di gruppi di istanze gestite
che possono essere utilizzati per scalare automaticamente le risorse.
Per esaminare le metriche della GPU utilizzando Cloud Monitoring, completa
i seguenti passaggi:
Per assicurarti che tu e il service account della VM disponiate delle autorizzazioni
necessarie per monitorare le prestazioni della GPU sulle VM Windows,
chiedi all'amministratore di concedere a te e al service account della VM i
seguenti ruoli IAM nel progetto:
L'amministratore potrebbe anche assegnare a te e al service account della VM
le autorizzazioni richieste tramite ruoli
personalizzati o altri ruoli
predefiniti.
Configura lo script per la generazione di report sulle metriche GPU
Requisiti
In ogni VM, verifica che vengano soddisfatti i seguenti requisiti:
Apri un terminale PowerShell come amministratore e utilizza il
comando Invoke-WebRequest per scaricare lo script.
Invoke-WebRequest è disponibile su PowerShell 3.0 o versioni successive.
Google Cloud consiglia di utilizzare ctrl+v per incollare i blocchi di codice copiati.
mkdir c:\google-scripts
cd c:\google-scripts
Invoke-Webrequest -uri https://raw.githubusercontent.com/GoogleCloudPlatform/compute-gpu-monitoring/main/windows/gce-gpu-monitoring-cuda.ps1 -outfile gce-gpu-monitoring-cuda.ps1
Esegui lo script:
cd c:\google-scripts
.\gce-gpu-monitoring-cuda.ps1
Configura l'agente in modo che si avvii automaticamente insieme al sistema
Per assicurarti che l'agente per la generazione di report sulle metriche GPU sia configurato in modo da iniziare l'esecuzione
all'avvio del sistema, utilizza il seguente comando per aggiungerlo a Utilità di pianificazione
di Windows.
$Trigger= New-ScheduledTaskTrigger -AtStartup
$Trigger.ExecutionTimeLimit = "PT0S"
$User= "NT AUTHORITY\SYSTEM"
$Action= New-ScheduledTaskAction -Execute "PowerShell.exe" -Argument "C:\google-scripts\gce-gpu-monitoring-cuda.ps1"
$settingsSet = New-ScheduledTaskSettingsSet
# Set the Execution Time Limit to unlimited on all versions of Windows Server
$settingsSet.ExecutionTimeLimit = 'PT0S'
Register-ScheduledTask -TaskName "MonitoringGPUs" -Trigger $Trigger -User $User -Action $Action -Force -Settings $settingsSet
Esamina le metriche in Cloud Monitoring
Nella console Google Cloud , vai alla pagina Esplora metriche.
[[["Facile da capire","easyToUnderstand","thumb-up"],["Il problema è stato risolto","solvedMyProblem","thumb-up"],["Altra","otherUp","thumb-up"]],[["Difficile da capire","hardToUnderstand","thumb-down"],["Informazioni o codice di esempio errati","incorrectInformationOrSampleCode","thumb-down"],["Mancano le informazioni o gli esempi di cui ho bisogno","missingTheInformationSamplesINeed","thumb-down"],["Problema di traduzione","translationIssue","thumb-down"],["Altra","otherDown","thumb-down"]],["Ultimo aggiornamento 2025-07-15 UTC."],[[["\u003cp\u003eThis document guides you through tracking GPU usage rates on Windows virtual machines (VMs) to optimize resource utilization, including setting up managed instance groups for autoscaling.\u003c/p\u003e\n"],["\u003cp\u003eMonitoring GPU performance requires granting specific Identity and Access Management (IAM) roles, such as Compute Instance Admin (v1) and Monitoring Metric Writer, to both the VM's service account and your user account.\u003c/p\u003e\n"],["\u003cp\u003eTo monitor GPU metrics, you must set up and run a GPU metrics reporting script on each VM, ensuring each VM has GPUs attached and a GPU driver installed.\u003c/p\u003e\n"],["\u003cp\u003eThe GPU metrics reporting agent can be configured to automatically start on boot via the Windows Task Scheduler, ensuring continuous data collection.\u003c/p\u003e\n"],["\u003cp\u003eYou can view collected GPU data, such as utilization, memory usage, and temperature, in Google Cloud's Metrics Explorer by selecting the custom metrics under the VM Instance resource.\u003c/p\u003e\n"]]],[],null,["# Monitoring GPU performance on Windows VMs\n\nWindows\n\n*** ** * ** ***\n\nTo help with better utilization of resources, you can track the GPU usage\nrates of your virtual machine (VM) instances.\n\nWhen you know the GPU usage rates,\nyou can perform tasks such as\nsetting up [managed instance groups](/compute/docs/instance-groups#managed_instance_groups)\nthat can be used to autoscale resources.\n\nTo review GPU metrics using [Cloud Monitoring](/monitoring/docs), complete\nthe following steps:\n\n1. On each VM, [set up the GPU metrics reporting script](#setup-script). This script installs the GPU metrics reporting agent. This agent runs at intervals on the VM to collect GPU data, and sends this data to Cloud Monitoring.\n2. On each VM, [run the script](#run-script).\n3. On each VM, set GPU metrics reporting agent to [automatically start on boot](#autostart).\n4. [View logs in Google Cloud Cloud Monitoring](#review-metrics).\n\nRequired roles\n--------------\n\nTo monitor GPU performance on Windows VMs, you need to grant the required\nIdentity and Access Management (IAM) roles to the following principles:\n\n- The [service account](/compute/docs/instances/change-service-account#serviceaccountname) that is used by the VM instance\n- Your user account\n\n\nTo ensure that you and the VM's service account has the necessary\npermissions to monitor GPU performance on Windows VMs,\n\nask your administrator to grant you and the VM's service account the\nfollowing IAM roles on the project:\n\n- [Compute Instance Admin (v1)](/iam/docs/roles-permissions/compute#compute.instanceAdmin.v1) (`roles/compute.instanceAdmin.v1`)\n- [Monitoring Metric Writer](/iam/docs/roles-permissions/monitoring#monitoring.metricWriter) (`roles/monitoring.metricWriter`)\n\n\nFor more information about granting roles, see [Manage access to projects, folders, and organizations](/iam/docs/granting-changing-revoking-access).\n\n\nYour administrator might also be able to give you and the VM's service account\nthe required permissions through [custom\nroles](/iam/docs/creating-custom-roles) or other [predefined\nroles](/iam/docs/roles-overview#predefined).\n\nSet up the GPU metrics reporting script\n---------------------------------------\n\n### Requirements\n\nOn each of your VMs, check that you meet the following requirements:\n\n- Each VM must have [GPUs attached](/compute/docs/gpus/create-vm-with-gpus).\n- Each VM must have a [GPU driver installed](/compute/docs/gpus/install-drivers-gpu#verify-driver-install).\n\n### Download the script\n\nOpen a PowerShell terminal as an administrator and use the\n[`Invoke-WebRequest` command](https://docs.microsoft.com/en-us/powershell/module/microsoft.powershell.utility/invoke-webrequest?view=powershell-7.1) to download the script.\n\n`Invoke-WebRequest` is available on PowerShell 3.0 or later.\nGoogle Cloud recommends that you use `ctrl+v` to paste the copied code blocks. \n\n```bash\nmkdir c:\\google-scripts\ncd c:\\google-scripts\nInvoke-Webrequest -uri https://raw.githubusercontent.com/GoogleCloudPlatform/compute-gpu-monitoring/main/windows/gce-gpu-monitoring-cuda.ps1 -outfile gce-gpu-monitoring-cuda.ps1\n```\n\nRun the script\n--------------\n\n```bash\ncd c:\\google-scripts\n.\\gce-gpu-monitoring-cuda.ps1\n```\n\nConfigure the agent to automatically start on boot\n--------------------------------------------------\n\nTo ensure that the GPU metrics reporting agent agent is set up to run on\nsystem boot, use the following command to add the agent to the Windows Task\nScheduler. \n\n```bash\n$Trigger= New-ScheduledTaskTrigger -AtStartup\n$Trigger.ExecutionTimeLimit = \"PT0S\"\n$User= \"NT AUTHORITY\\SYSTEM\"\n$Action= New-ScheduledTaskAction -Execute \"PowerShell.exe\" -Argument \"C:\\google-scripts\\gce-gpu-monitoring-cuda.ps1\"\n$settingsSet = New-ScheduledTaskSettingsSet\n# Set the Execution Time Limit to unlimited on all versions of Windows Server\n$settingsSet.ExecutionTimeLimit = 'PT0S'\nRegister-ScheduledTask -TaskName \"MonitoringGPUs\" -Trigger $Trigger -User $User -Action $Action -Force -Settings $settingsSet\n```\n\nReview metrics in Cloud Monitoring\n----------------------------------\n\n1. In the Google Cloud console, go to the **Metrics Explorer** page.\n\n [Go to Monitoring](https://console.cloud.google.com/monitoring/metrics-explorer)\n2. Expand the **Select a metric** menu.\n\n3. In the **Resource** menu, select **VM Instance**.\n\n4. In the **Metric category** menu, select **Custom**.\n\n5. In the **Metric** menu, select the metric to chart. For example\n `custom/instance/gpu/utilization`.\n\n | **Note:** Custom metrics might take some time to display.\n6. Click **Apply**.\n\n Your GPU utilization should resemble the following output:\n\nAvailable metrics\n-----------------\n\nWhat's next?\n------------\n\n- To handle GPU host maintenance, see [Handling GPU host maintenance events](/compute/docs/gpus/gpu-host-maintenance).\n- To improve network performance, see [Use higher network bandwidth](/compute/docs/gpus/optimize-gpus)."]]
