Standard 模式:您需要为每个节点上的计算资源付费,而无论 Pod 是否在该节点上运行。您需要为未使用的资源付费,因此您应该管理工作负载调度,以尽量减少节点中的资源浪费。如需了解详情,请参阅 Standard 价格。
确保 Standard 集群中的资源使用效率水平一致需要您持续监控集群的状态。在 Autopilot 集群中,GKE 会为您执行监控和管理。
在 Standard 集群中,您需要为节点上未使用的计算资源付费。您可以通过装箱来降低这些费用,在该方法中,您需要将尽可能多的 Pod 放在每个节点上,以避免浪费容量。装箱需要持续的工作负载管理和调度自定义。使用 Autopilot 集群时,您无需对工作负载进行装箱,因为您只需为工作负载使用的资源付费。
何时使用 Standard 而非 Autopilot
虽然我们建议为大多数工作负载使用 Autopilot,但是由于预配置的安全强化或默认集群配置,Autopilot 可能无法满足特定要求。在以下情况下,您应该考虑使用 Standard 模式,而非 Autopilot 模式:
[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-08-01。"],[],[],null,["# Choose a GKE mode of operation\n\n[Autopilot](/kubernetes-engine/docs/concepts/autopilot-overview) [Standard](/kubernetes-engine/docs/concepts/choose-cluster-mode)\n\n*** ** * ** ***\n\nThis page helps you to choose the Google Kubernetes Engine (GKE) *mode of operation*\nthat's the best fit for your workloads. This page is intended for platform\nadministrators who are considering GKE as a managed Kubernetes\nprovider and want to discover the options available in Google Cloud. If\nyou want to learn about whether GKE as a *platform* is the best\nchoice for your containerized applications, refer to the\n[GKE Overview](/kubernetes-engine/docs/concepts/kubernetes-engine-overview) and [GKE and Cloud Run](/kubernetes-engine/docs/concepts/gke-and-cloud-run).\n\nGKE offers the following modes of operation for clusters:\n\n- **Autopilot mode (recommended)**: GKE manages the underlying infrastructure such as node configuration, autoscaling, auto-upgrades, baseline security configurations, and baseline networking configuration.\n- **Standard mode**: You manage the underlying infrastructure, including configuring the individual nodes.\n\nYou can't convert a cluster from Standard to Autopilot after\ncluster creation. We recommend that you read this page and, optionally, read the\n[Autopilot and Standard comparison](/kubernetes-engine/docs/resources/autopilot-standard-feature-comparison)\nso that you can make an informed choice.\n\nWhy GKE Autopilot mode\n----------------------\n\nGoogle manages most of the infrastructure in a GKE\nAutopilot cluster, providing a more *managed* Kubernetes experience\nthan GKE Standard mode. The default configuration of\nAutopilot clusters is optimized for most production workloads.\nGKE Autopilot implements many Kubernetes best practices\nfor security, scalability, and cost optimization by default.\n\n**In most cases, we recommend running your production workloads on\nAutopilot.**\n\nAutopilot provides a default configuration that has benefits such as\nthe following:\n\n- **Cost efficiency**: You only pay for the compute resources that your workloads use while running. You don't pay for unused capacity on your nodes, system Pods, operating system costs, or unscheduled workloads.\n- **Automation**: Google manages nodes, creating new nodes for your apps and configuring automatic upgrades and repairs. GKE automatically scales nodes and workloads based on traffic.\n- **Improved security posture and reliability**: Autopilot clusters enable many GKE security settings and Kubernetes best practices by default. GKE automatically applies security patches to your nodes when available.\n\nFor a full list of benefits in GKE Autopilot, refer to\n[About GKE Autopilot](/kubernetes-engine/docs/concepts/autopilot-overview).\n\nWhy GKE Standard mode\n---------------------\n\nIn Standard mode, you manage every configuration setting in your\ncluster and nodes, including managing groups of nodes called *node pools* that\nshare characteristics. Under the\n[shared responsibility model](/kubernetes-engine/docs/concepts/shared-responsibility),\nGoogle still manages your control plane, but you must configure your nodes.\nSettings that you manage yourself include the following:\n\n- **Node pools**: You create and manage groups of nodes that have similar configuration settings.\n- **Security** : GKE Standard clusters have default hardening measures applied, but many GKE security features are not enabled by default, such as [Workload Identity Federation for GKE](/kubernetes-engine/docs/concepts/workload-identity) and [Shielded GKE Nodes](/kubernetes-engine/docs/how-to/shielded-gke-nodes). You can enable these features manually and configure the settings.\n- **Scheduling** : You must monitor and design your workloads so that GKE can schedule them efficiently on your nodes to minimize unused resources (*bin-packing*).\n- **Scaling** : You must set up and configure [node auto-provisioning](/kubernetes-engine/docs/how-to/node-auto-provisioning), configure [automatic scaling](/kubernetes-engine/docs/concepts/cluster-autoscaler) settings, and ensure that your nodes don't have too many resources or too few resources.\n- **Resource management**: You must evaluate the resource needs of each workload that you run on Standard clusters to ensure that the resource requests meet the workload requirements.\n- **Version management**: Best practices such as automatic GKE version upgrades and release channel enrollment are off by default in Standard. You can configure auto-upgrades and GKE versions when you create or update the cluster.\n\nPricing differences\n-------------------\n\nThe pricing model for Autopilot is different from Standard, as\nfollows:\n\n- **Autopilot mode** : You only pay for the compute resources that your workloads use while running. You don't pay for unused resources on nodes, OS running costs, unscheduled workloads, or system workloads. For details, refer to [Autopilot pricing](/kubernetes-engine/pricing#autopilot_mode).\n- **Standard mode** : You pay for the compute resources on each node, regardless of whether Pods run on the node. You pay for unused resources, so you should manage workload scheduling to minimize resource wastage in nodes. For details, refer to [Standard pricing](/kubernetes-engine/pricing#standard_mode).\n\nEnsuring a consistent level of resource usage efficiency in Standard\nclusters requires that you constantly monitor the state of your cluster. In\nAutopilot clusters, GKE does the monitoring and\nmanagement for you.\n\nIn Standard clusters, you pay for unused compute resources on your\nnodes. You can reduce these costs by\n[*bin-packing*](/blog/products/containers-kubernetes/gke-best-practices-to-lessen-over-provisioning),\nin which you place as many Pods as possible onto each node to avoid wasted\ncapacity. Bin-packing requires constant workload management and scheduling\ncustomization. Autopilot clusters eliminate the need for you to\nbin-pack your workloads because you only pay for the resources that your\nworkloads use.\n\nWhen to use Standard instead of Autopilot\n-----------------------------------------\n\nWhile we recommend using Autopilot for most workloads,\nyou might have specific requirements that Autopilot can't meet due to\nthe pre-configured hardening or the default cluster configuration. You should\nconsider using Standard mode over Autopilot mode in the\nfollowing scenarios:\n\n- You require granular control over your cluster and node configuration,\n including the ability to directly connect to your nodes using SSH.\n\n | **Note:** You can still get shell access to your containers in Autopilot clusters that have the `sh` binary, but not your nodes, because GKE manages Autopilot nodes. We recommend that you disable node SSH access in GKE Standard unless required.\n- You want to install or modify software running on the nodes themselves, such\n as changing the node operating system.\n\n- You want to\n [customize the node system configuration](/kubernetes-engine/docs/how-to/node-system-config),\n such as by setting\n [Linux sysctls](https://www.kernel.org/doc/Documentation/sysctl/net.txt).\n\n- You want to perform actions that Autopilot restricts, for example,\n running workloads in GKE-managed namespaces such as\n `kube-system`. We recommend that you don't deploy workloads in these\n namespaces.\n\n- You want to use specific GKE features that are only available\n on Standard, such as\n [Cloud TPU](/kubernetes-engine/docs/concepts/tpus).\n\n- You want to test alpha features in open source Kubernetes.\n\n- You want to provision additional unused capacity in your cluster.\n\nUnless you have specific requirements such as these, we recommend that you try\nAutopilot for your workloads. For an interactive walkthrough that sets\nup an Autopilot cluster and exposes a `hello-world` application, go to\nthe Autopilot walkthrough in the Google Cloud console:\n\n[Go to walkthrough](https://console.cloud.google.com/welcome?walkthrough_id=gke_autopilot)\n\nWhat's next\n-----------\n\n- [View a detailed capability comparison between Autopilot and Standard](/kubernetes-engine/docs/resources/autopilot-standard-feature-comparison).\n- [Create an Autopilot cluster](/kubernetes-engine/docs/how-to/creating-an-autopilot-cluster).\n- [Learn about GKE cluster architecture](/kubernetes-engine/docs/concepts/cluster-architecture).\n- [Learn more about Autopilot](/kubernetes-engine/docs/concepts/autopilot-overview)."]]