updated Provisioned Control Plane docs based on recent feedback.

Author: georgejohnis

latest/ug/clusters/eks-provisioned-control-plane-introduction.adoc

@@ -7,34 +7,32 @@ include::../attributes.txt[]
 
 [abstract]
 --
-Learn how to use Amazon EKS Provisioned Control Plane to guarantee minimum control plane capacity for predictable performance.
+Learn how to use Amazon EKS Provisioned Control Plane for predictable and high performance from cluster's control plane.
 --
 
 == Overview
 
 Amazon EKS Provisioned Control Plane is a feature that enables cluster
-administrators to select from a set of control plane scaling tiers and
-designate their chosen tier as the guaranteed minimum control plane
-capacity. This enables cluster administrators to ensure that the
-cluster's control plane is always provisioned with the specified minimum
+administrators to select from a set of scaling tiers and
+designate their chosen tier for high, predictale performance from the cluster's control plane. This enables cluster administrators to ensure that the
+cluster's control plane is always provisioned with the specified 
 capacity, regardless of actual utilization. Each control plane scaling
 tier offers significantly increased compute capacity and performance
 compared to what was available previously.
 
 By default, Amazon EKS cluster control planes scale up and down
-automatically in response to a comprehensive set of signals, including
-control plane instance utilization and etcd database size. This is the
+automatically in response to a  signals such as 
+control plane utilization. This is the
 *standard mode* of EKS Control Plane operation and works well for a wide
 range of use cases by dynamically allocating sufficient capacity as
 needed. However, for workloads characterized by unpredictable traffic
 spikes, extreme bursts, or those requiring very high amounts of control
 plane capacity, standard mode may not be ideal. EKS Provisioned Control
-Plane is the right solution for workloads that require a guaranteed
-minimum amount of control plane capacity that is instantly available.
+Plane is the right solution for workloads that require predictable, high performance from the control plane that is instantly available.
 
 With EKS Provisioned Control Plane, cluster administrators can
 pre-provision desired control plane capacity ahead of time, providing
-predictable performance from the cluster's control plane even in
+predictable and high performance from the cluster's control plane even in
 scenarios with bursty or erratic traffic patterns. Additionally, EKS
 Provisioned Control Plane enables cluster administrators to provision
 the same control plane capacity across environments—from development and
@@ -47,7 +45,7 @@ scalable AI workloads, high-performance computing, and large-scale data
 processing workloads on Kubernetes.
 
 All existing and new Amazon EKS clusters operate in standard mode by
-default. For clusters requiring guaranteed capacity for predictable
+default. For clusters requiring  predictable and high 
 performance from the control plane, you can opt in to use the EKS
 Provisioned Control Plane feature. You will be billed at the hourly rate
 for the particular control plane scaling tier in addition to the
@@ -60,11 +58,18 @@ image::images/control-plane-modes.png[Amazon EKS Control Plane Modes,scaledwidth
 == Use cases
 
 EKS Provisioned Control Plane is designed to address specific scenarios
-where guaranteed control plane capacity and predictable performance are
+where predictable and high control plane performance are
 critical to your operations. Understanding these use cases can help you
 determine whether EKS Provisioned Control Plane is the right solution
 for your workloads.
 
+*Performance-critical workloads* – For workloads that demand minimal
+latency and maximum performance from the Kubernetes control plane, EKS
+Provisioned Control Plane provides capacity that eliminates
+the variability inherent in reactive scaling. This is particularly
+important for applications where control plane responsiveness directly
+impacts user experience or business outcomes.
+
 *Massively scalable workloads* – If you run highly scalable workloads
 such as AI training and inference, high-performance computing, or
 large-scale data processing that require large number of nodes running
@@ -82,13 +87,6 @@ you to scale your control plane capacity in advance. This proactive
 approach ensures your control plane is ready to handle the increased
 load without waiting for automatic scaling to respond to demand.
 
-*Performance-critical workloads* – For workloads that demand minimal
-latency and maximum performance from the Kubernetes control plane, EKS
-Provisioned Control Plane provides guaranteed capacity that eliminates
-the variability inherent in reactive scaling. This is particularly
-important for applications where control plane responsiveness directly
-impacts user experience or business outcomes.
-
 *Environment consistency* – EKS Provisioned Control Plane enables you to
 match control plane capacity and performance across development and
 production environments, helping you identify potential issues early
@@ -129,10 +127,10 @@ available in the etcd database used to store cluster state, measured in
 gigabytes. This attribute determines how many Kubernetes objects (pods,
 services, deployments, etc.) your cluster can store.
 
-When you provision your cluster's control plane on a certain tier using
+When you provision your cluster's control plane on a certain scaling tier using
 EKS Provisioned Control Plane, Amazon EKS ensures your cluster's control
-plane maintains the capacity level corresponding to that tier. The
-capabilities of control plane scaling tiers vary by Kubernetes version,
+plane maintains the limits corresponding to that tier. The
+limits of control plane scaling tiers vary by Kubernetes version,
 as shown in the following tables.
 
 === EKS v1.28 and v1.29
@@ -162,7 +160,6 @@ as shown in the following tables.
 
 === EKS v1.30 and later
 
-
 [cols="1,1,1,1",options="header",]
 |===
 |Provisioned Control Plane Scaling Tier 
@@ -186,15 +183,12 @@ as shown in the following tables.
 |18
 |===
 
-
 === Monitoring control plane scaling tier utilization
 
 Amazon EKS provides several metrics to help you monitor your control
-plane's tier utilization. These metrics are published as Amazon CloudWatch  metrics and are
+plane's tier utilization. These metrics are published as <<cloudwatch,Amazon CloudWatch metrics>> and are
 accessible through the CloudWatch and EKS console. Additionally, these metrics
-are scrapable from your EKS cluster's Prometheus endpoint (see
-https://docs.aws.amazon.com/eks/latest/userguide/view-raw-metrics.html[here]).
-
+are scrapable from your EKS cluster's Prometheus endpoint (see <<prometheus,here>>
 
 
 [cols="1,1,1",options="header",]
@@ -214,22 +208,16 @@ https://docs.aws.amazon.com/eks/latest/userguide/view-raw-metrics.html[here]).
 
 
 |*Cluster database size*
-|apiserver_storage_size_bytes, etcd_mvcc_db_total_size_in_use_in_bytes*
-|apiserver_storage_size_bytes, etcd_mvcc_db_total_size_in_use_in_bytes*
+|apiserver_storage_size_bytes, etcd_mvcc_db_total_size_in_use_in_bytes
+|apiserver_storage_size_bytes, etcd_mvcc_db_total_size_in_use_in_bytes
 |===
 
 
 NOTE: The etcd_mvcc_db_total_size_in_use_in_bytes metric is currently being rolled out gradually across all EKS clusters. The metric is  expected to be available in all clusters by the end of Q2 2026
 
-=== Selecting the optimal tier
+=== Understanding Tier capacity versus actual performance
 
-To determine the optimal tier for your cluster, you can perform load
-testing by provisioning your Amazon EKS cluster in EKS Provisioned
-Control Plane mode using the highest tier (4XL). Then perform a load
-test to simulate peak demand on your cluster's control plane. Observe
-the control plane tier utilization metrics at peak load, and use these
-observations as the guiding factor to select the appropriate tier for
-provisioning capacity.
+When you select a Provisioned Control Plane scaling tier, the tier attributes represent the underlying configurations that Amazon EKS applies to your control plane. However, the actual performance you achieve depends on your specific workload patterns, configurations, and adherence to Kubernetes best practices. For example, while a 4XL tier configures API Priority and Fairness (APF) with 6,800 concurrent request seats, the actual request throughput you obtain from the control plane depends on the types of operations being performed. For example, list operations consume more resources than get operations, and cluster-wide queries are more expensive than namespace-scoped queries (see https://docs.aws.amazon.com/eks/latest/best-practices/scale-control-plane.html#_api_priority_and_fairness[here]). Similarly, although the default scheduler QPS is set to 400 for a 4XL tier, your actual pod scheduling rate depends on factors like node readiness and health. To achieve optimal performance, ensure your applications follow Kubernetes best practices and are properly configured for your workload characteristics.
 
 == Considerations
 
@@ -238,13 +226,12 @@ uses a flexible fleet of instances in the backend to offer the best
 price to performance ratio. However this means that the compute
 capabilities delivered can vary, so the actual performance limits are
 not fixed and can fluctuate. While these values can occasionally scale
-to levels comparable to the lower EKS Provisioned Control Plane tier
-(XL), they are not guaranteed. If your workload requires consistent,
-predictable control plane performance with clearly defined minimum
+to levels comparable to the lower EKS Provisioned Control Plane tiers, they may not be available in all clluster. If your workload requires predictable, high 
+control plane performance with clearly defined minimum
 capabilities, you should consider upgrading to an EKS Provisioned
-Control Plane tier where these metrics are explicitly guaranteed.
+Control Plane tier.
 * *Opt-in required* – Existing clusters will not automatically scale up
-from the standard control plane to a higher-priced EKS Provisioned
+from the standard control plane to a higher https://aws.amazon.com/eks/pricing/[priced] EKS Provisioned
 Control Plane tier. You must explicitly opt in to one of the new EKS
 Provisioned Control Plane scaling tiers.
 * *Exiting restriction* – You cannot exit an EKS Provisioned Control
@@ -278,5 +265,16 @@ occurs. The error field in the update indicates the reason for failure.
 There are no restrictions on how frequently you can switch between
 tiers. Changing the control plane tier takes several minutes to
 complete.
+* *Selecting optimal tier* – To determine the optimal tier for your cluster, you can perform load testing by provisioning your Amazon EKS cluster in EKS Provisioned Control Plane mode using the highest tier (4XL). Then perform a load
+test to simulate peak demand on your cluster's control plane. Observe
+the control plane tier utilization metrics at peak load, and use these
+observations as the guiding factor to select the appropriate tier for
+provisioning capacity.
+* *Provisioned Control Plane pricing* – You will be billed at the hourly rate for the Provisioned Control Plane scaling tier your cluster is on. This is in addition to the standard or extended support hourly charges. See Amazon EKS Pricing https://aws.amazon.com/eks/pricing/[page] for details.
+* *Larger scaling tier* – If you intend to run your cluster on scaling tier larger than 4XL, contact your AWS account team for additional pricing information.
+* *Kuberentes version and region support* – EKS Provisioned Control Plane is supported in all AWS commercial, GovCloud, and China regions. Provisioned Control Plane works on EKS v1.28 and higher.
+
+
+
 
 include::eks-provisioned-control-plane-getting-started.adoc[leveloffset=+1]

latest/ug/images/control-plane-modes.png

@@ -27,6 +27,14 @@ For clusters that are Kubernetes version `1.28` and above, you get CloudWatch ve
 |Metric name
 |Description
 
+|`apiserver_flowcontrol_current_executing_seats`
+| The number of seats currently in use for executing API requests. Seat allocation is determined by the priority_level and flow_schema configuration in the Kubernetes API Priority and Fairness 
+https://kubernetes.io/docs/concepts/cluster-administration/flow-control/[feature].
+
+*Units:* Count
+
+*Valid statistics:* Max
+
 |`scheduler_schedule_attempts_total`
 |The number of total attempts by the scheduler to schedule Pods in the cluster for a given period. This metric helps monitor the scheduler's workload and can indicate scheduling pressure or potential issues with Pod placement.