注意:請為支援 Private Service Connect (PSC) 的 Vertex AI 服務使用這項功能,因為這是建議的連線方法。只有在無法使用 PSC 的服務,才使用私人服務存取權。
在「Vertex AI 的私密存取選項」表格的「私人服務連線」欄中,凡是標有勾號的 Vertex AI 服務,都必須透過私人服務連線連線。
這些 Google 管理的 Vertex AI 服務支援與服務取用者內部部署、多雲端和 VPC 工作負載的雙向通訊。
這項私密通訊完全使用內部 IP 位址。VM 執行個體不需要網際網路存取權或外部 IP 位址,就可以透過私人服務存取權與服務連線。
Vertex AI 提供的服務會代管在 Google 管理的 VPC 網路中。透過私人服務存取權,您可以使用虛擬私有雲網路對等互連連線,連線至這些 Vertex AI 和第三方服務的內部 IP 位址。
下圖顯示自訂訓練架構,其中訓練工作和管道工作的 Vertex AI API 會在服務專案 (serviceproject) 中啟用及管理,做為共用虛擬私有雲部署作業的一部分。這些元件會以 Google 管理的基礎架構即服務 (IaaS) 形式,部署在服務生產者的 VPC 網路中。服務用戶的虛擬私有雲網路 (hostproject) 會透過私人服務存取連線存取這些服務。
私人服務存取權部署選項
您可以建立新的私人連線,或修改現有連線。
設定私人服務存取權之前,請先瞭解選擇虛擬私有雲網路和 IP 位址範圍的注意事項。
如要建立新的私人連線,您必須先建立已分配的 IP 範圍,然後在虛擬私有雲網路與 Google 管理的 Vertex AI 服務之間建立私人連線。
[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間:2025-09-04 (世界標準時間)。"],[],[],null,["# About accessing Vertex AI services through private services access\n\nVertex AI services that have a checkmark\nin the **Private services access**\ncolumn of the\n[Private access options for Vertex AI](/vertex-ai/docs/general/netsec-overview#private_access_options_for)\ntable\nrequire you to connect to their services through\n[private services access](/vpc/docs/configure-private-services-access).\n\nThese Google-managed Vertex AI services support bidirectional\ncommunication with a service consumer's on-premises, multicloud, and\nVPC workloads.\n\nThis private communication happens exclusively by using internal\nIP addresses. VM instances don't need internet access or external IP\naddresses to reach services that are available through private services access.\n\nVertex AI provides services that are hosted in a Google-managed\nVPC network. Private services access lets you reach the\ninternal IP addresses of these Vertex AI and third-party services\nthrough a VPC Network Peering connection.\n\nThe following diagram shows a\n[custom training](/vertex-ai/docs/training/overview)\narchitecture in which Vertex AI APIs for training jobs and\npipeline jobs are enabled and managed in a service project\n(`serviceproject`) as part of a [Shared VPC](/vpc/docs/shared-vpc)\ndeployment.\nThese components are deployed as a Google-managed\nInfrastructure-as-a-Service (IaaS) in the service producer's\nVPC network.\nThe service consumer's VPC network (`hostproject`) accesses\nthese services through a private services access connection.\n\nPrivate services access deployment options\n------------------------------------------\n\nYou can create a new private connection or modify an existing one.\nBefore you configure private services access, understand the\n[considerations](/vpc/docs/configure-private-services-access#considerations)\nfor choosing a VPC network and IP address range.\n\nTo create a new private connection, you must first create an\n[allocated IP range](/vpc/docs/configure-private-services-access#procedure)\nand then create a\n[private connection](/vpc/docs/configure-private-services-access#creating-connection)\nbetween your VPC network and\nGoogle-managed Vertex AI services.\n\nAlternatively, you can modify an existing connection. For more information, see\n[Modify a private connection](/vpc/docs/configure-private-services-access#modifying-connection).\n\n### Vertex AI subnet recommendations\n\nThe following table lists the recommended subnet ranges for Vertex AI\nservices.\n\nDeployment considerations\n-------------------------\n\nFollowing are some important considerations that affect how you establish\ncommunication between your\non-premises, multicloud, and VPC workloads and Google-managed\nVertex AI services.\n\n### IP advertisement\n\nYou must advertise the private services access subnet range from the\nCloud Router as a custom advertised route. For more information, see\n[Advertise custom IP ranges](/network-connectivity/docs/router/how-to/advertising-custom-ip).\n\n### VPC Network Peering\n\nThe service producer's network might not have the correct routes to direct\ntraffic to your on-premises network. By default, the service producer's\nnetwork only learns the subnet routes from your VPC network. Therefore,\nany request that's not from a subnet IP range is dropped by the\nservice producer.\n\nFor this reason, in your VPC network, you must\n[update the peering connection](/vpc/docs/using-vpc-peering#update-peer-connection)\nto export custom routes to the service producer's network. Exporting routes sends all\neligible static and dynamic routes that are in your VPC network, such as\nroutes to your on-premises network, to the service producer's network.\nThe service producer's network automatically imports them and then can send\ntraffic back to your on-premises network through the VPC network.\n\n### Firewall rules\n\nYou must update the firewall rules for the VPC\nnetwork that connects your on-premises and multicloud environments\nto Google Cloud to allow ingress traffic from and egress traffic to private\nservices access subnets."]]