이 페이지에서는 Arm VM을 일괄 및 스트리밍 Dataflow 작업의 작업자로 사용하는 방법을 설명합니다.
Arm 프로세서의 Tau T2A 머신 시리즈 및 C4A 머신 시리즈(프리뷰)를 사용하여 Dataflow 작업을 실행할 수 있습니다. Arm 아키텍처는 전력 효율성에 최적화되어 있으므로 이러한 VM을 사용하면 일부 워크로드의 가성비가 향상됩니다. Arm VM에 대한 자세한 내용은 Compute의 Arm VM을 참조하세요.
요구사항
다음 Apache Beam SDK는 Arm VM을 지원합니다.
Apache Beam Java SDK 버전 2.50.0 이상
Apache Beam Python SDK 버전 2.50.0 이상
Apache Beam Go SDK 버전 2.50.0 이상
Tau T2A 또는 C4A 머신을 사용할 수 있는 리전을 선택합니다. 자세한 내용은 사용 가능한 리전 및 영역을 참조하세요.
[[["이해하기 쉬움","easyToUnderstand","thumb-up"],["문제가 해결됨","solvedMyProblem","thumb-up"],["기타","otherUp","thumb-up"]],[["이해하기 어려움","hardToUnderstand","thumb-down"],["잘못된 정보 또는 샘플 코드","incorrectInformationOrSampleCode","thumb-down"],["필요한 정보/샘플이 없음","missingTheInformationSamplesINeed","thumb-down"],["번역 문제","translationIssue","thumb-down"],["기타","otherDown","thumb-down"]],["최종 업데이트: 2025-09-04(UTC)"],[[["\u003cp\u003eArm VMs, including Tau T2A and C4A machine series, can be used as workers for Dataflow batch and streaming jobs, offering improved price-performance for certain workloads due to their power efficiency.\u003c/p\u003e\n"],["\u003cp\u003eArm VM support requires specific Apache Beam SDK versions (2.50.0 or later for Java, Python, and Go), availability in select regions, use of Runner v2, and Streaming Engine for streaming jobs.\u003c/p\u003e\n"],["\u003cp\u003eRunning Dataflow jobs on Arm VMs requires setting the \u003ccode\u003eworkerMachineType\u003c/code\u003e (Java) or \u003ccode\u003emachine_type\u003c/code\u003e/\u003ccode\u003eworker_machine_type\u003c/code\u003e (Python/Go) pipeline option and specifying an ARM machine type.\u003c/p\u003e\n"],["\u003cp\u003eThere are several limitations to consider, such as unsupported GPUs, Cloud Profiler, Dataflow Prime, worker VM metrics, and container image pre-building, in addition to the limitations that also apply to Tau T2A and C4A machines.\u003c/p\u003e\n"],["\u003cp\u003eUsing custom containers require multi-architecture images, to ensure they match the architecture of the worker VMs.\u003c/p\u003e\n"]]],[],null,["# Use Arm VMs on Dataflow\n\nThis page explains how to use Arm VMs as workers for batch and streaming\nDataflow jobs.\n\nYou can use the\n[Tau T2A machine series](/compute/docs/general-purpose-machines#t2a_machines)\nand [C4A machine series](/compute/docs/general-purpose-machines#c4a_series)\n([Preview](/products#product-launch-stages)) of\nArm processors to run Dataflow jobs. Because Arm architecture is\noptimized for power efficiency, using these VMs yields better price for\nperformance for some workloads. For more information about Arm VMs, see\n[Arm VMs on Compute](/compute/docs/instances/arm-on-compute).\n\nRequirements\n------------\n\n- The following Apache Beam SDKs support Arm VMs:\n - Apache Beam Java SDK versions 2.50.0 or later\n - Apache Beam Python SDK versions 2.50.0 or later\n - Apache Beam Go SDK versions 2.50.0 or later\n- Select a region where Tau T2A or C4A machines are available. For more information, see [Available regions and zones](/compute/docs/regions-zones#available).\n- Use [Runner v2](/dataflow/docs/runner-v2) to run the job.\n- Streaming jobs must use [Streaming Engine](/dataflow/docs/streaming-engine).\n\nLimitations\n-----------\n\n- All [Tau T2A limitations](/compute/docs/general-purpose-machines#t2a_limitations) and [C4A limitations](/compute/docs/general-purpose-machines#supported_disk_types_for_c4a) apply.\n- [GPUs](/dataflow/docs/gpu) are not supported.\n- [Cloud Profiler](/dataflow/docs/guides/profiling-a-pipeline) is not supported.\n- [Dataflow Prime](/dataflow/docs/guides/enable-dataflow-prime) is not supported.\n- Receiving worker VM metrics from [Cloud Monitoring](/dataflow/docs/guides/using-cloud-monitoring#receive_worker_vm_metrics_from_the_agent) is not supported.\n- [Container image pre-building](/dataflow/docs/guides/build-container-image#prebuild) is not supported.\n\nRun a job using Arm VMs\n-----------------------\n\nTo use Arm VMs, set the following pipeline option. \n\n### Java\n\nSet the `workerMachineType` pipeline option and specify an\n[ARM machine type](/compute/docs/instances/arm-on-compute).\n\nFor more information about setting pipeline options, see\n[Set Dataflow pipeline options](/dataflow/docs/guides/setting-pipeline-options).\n\n### Python\n\nSet the `machine_type` pipeline option and specify an\n[ARM machine type](/compute/docs/instances/arm-on-compute).\n\nFor more information about setting pipeline options, see\n[Set Dataflow pipeline options](/dataflow/docs/guides/setting-pipeline-options).\n\n### Go\n\nSet the `worker_machine_type` pipeline option and specify an\n[ARM machine type](/compute/docs/instances/arm-on-compute).\n\nFor more information about setting pipeline options, see\n[Set Dataflow pipeline options](/dataflow/docs/guides/setting-pipeline-options).\n\nUse multi-architecture container images\n---------------------------------------\n\nIf you use a custom container in Dataflow, the container must\nmatch the architecture of the worker VMs. If you plan to use a custom\ncontainer on ARM VMs, we recommend building a multi-architecture image. For more\ninformation, see\n[Build a multi-architecture container image](/dataflow/docs/guides/multi-architecture-container).\n\nPricing\n-------\n\nYou are billed for Dataflow compute resources.\nDataflow pricing is independent of the machine type family. For\nmore information, see [Dataflow pricing](/dataflow/pricing).\n\nWhat's next\n-----------\n\n- [Set Dataflow pipeline options](/dataflow/docs/guides/setting-pipeline-options)\n- [Use custom containers in Dataflow](/dataflow/docs/guides/using-custom-containers)"]]
