Container Request Right Sizing Recommendation API (v2)

The container request right sizing recommendation API provides recommendations for container resource requests based on configurable parameters and estimates the savings from implementing those recommendations on a per-container, per-controller level. Of course, if the cluster-level resources stay static then you will likely not enjoy real savings from applying these recommendations until you reduce your cluster resources. Instead, your idle allocation will increase.
The endpoint is available at


The algorithm to be used to calculate CPU recommendations based on historical CPU usage data. Options are max and quantile. Max recommendations are based on the maximum-observed usage in window. Quantile recommendations are based on a quantile of observed usage in window (requires the qCPU parameter to set the desired quantile). Defaults to max. NOTE: To use the quantile algorithm, the ContainerStats pipeline must be enabled. Please see the note on the .Values.kubecostModel.containerStatsEnabled value on the Helm chart.
Like algorithmCPU, but for RAM recommendations.
float in the range (0, 1]
The desired quantile to base CPU recommendations on. Only used if algorithmCPU=quantile. Note: a quantile of 0.95 is the same as a 95th percentile.
float in the range (0, 1]
Like qCPU, but for RAM recommendations.
float in the range (0,1]
An ratio of headroom on the base recommended CPU request. If the base recommendation is 100 mCPU and this parameter is 0.8, the recommended CPU request will be 100 / 0.8 = 125 mCPU. Defaults to 0.7. Inputs that fail to parse (see will default to 0.7.
float in the range (0,1]
Calculated like CPU.
Required parameter. Duration of time over which to calculate usage. Supports days before the current time in the following format: 3d. Note: Hourly windows are not currently supported. See the Allocation API documentation for more a more detailed explanation of valid inputs to window.
A filter to reduce the set of workloads for which recommendations will be calculated. See V2 Filters for syntax. V1 filters are also supported, please see v1 API documentation.

API examples

curl -G \
-d 'algorithmCPU=quantile' \
-d 'qCPU=0.95' \
-d 'algorithmRAM=max' \
-d 'targetCPUUtilization=0.8' \
-d 'targetRAMUtilization=0.8' \
-d 'window=3d' \
--data-urlencode 'filter=namespace:"kubecost"+container:"cost-model"' \

Recommendation methodology

The "base" recommendation is calculated from the observed usage of each resource per unique container spec (e.g. a 2-replica, 3-container Deployment will have 3 recommendations: one for each container spec).
Say you have a single-container Deployment with two replicas: A and B.
  • A's container had peak usages of 120 mCPU and 300 MiB of RAM.
  • B's container had peak usages of 800 mCPU and 120 MiB of RAM.
The max algorithm recommendation for the Deployment's container will be 800 mCPU and 300 MiB of RAM. Overhead will be added to the base recommendation according to the target utilization parameters as described above.

Savings projection methodology

See v1 docs.