About GKE Scalability

This page provides a set of recommendations for planning, architecting, deploying,scaling, and operating large workloads on Google Kubernetes Engine (GKE) clusters. We recommend you followthese recommendations to keep your scaling workloads withinservice-level objectives (SLOs).

Available recommendations for scalability

Before planning and designing a GKE architecture, map parameters specific to yourworkload (for example the number of active users, expected response time,required compute resources) with the resources used by Kubernetes (such as Pods,Services, and 'CustomResourceDefinition'). With this information mapped, reviewthe GKE scalability recommendations.

The scalability recommendations are divided based in the following planning scopes:

  • Plan for scalability: To learn about the general best practices fordesigning your workloads and clusters for reliable performance when runningon both small and large clusters. These recommendations are useful for architects,platform administrators, and Kubernetes developers. To learn more, seePlan for scalability.
  • Plan for large-size GKE clusters: To learn how to plan to run verybig-size GKE clusters. Learn about known limits of Kubernetes and GKE and waysto avoid reaching them. These recommendations are useful for architectsand platform administrators. To learn more, seePlan for large GKE clusters.
  • Plan for large workloads: To learn how to plan architectures that runlarge Kubernetes workloads on GKE. It covers recommendations for distributingthe workload among projects and clusters, and adjusting these workload requiredquotas. These recommendations are useful for architects and platform administrators.To learn more, seePlan for large workloads.

These scalability recommendations are general to GKE and are applicable to bothGKE Standard and GKE Autopilot modes. GKE Autopilot provisions and managesthe cluster's underlying infrastructure for you, therefore some recommendationsare not applicable.

Caution: Test your planned cluster configuration before its implementation.Some design decisions might include fixed parameters, for example, CIDRsdefinition. Changing these parameters on existing clusters is not availableand it requires cluster recreation.

What's next?

Except as otherwise noted, the content of this page is licensed under theCreative Commons Attribution 4.0 License, and code samples are licensed under theApache 2.0 License. For details, see theGoogle Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2026-02-18 UTC.