Kubernetes workloads often request more CPU and memory than they actually use, leading to wasted resources and inflated costs. CloudZero helps you identify this over-provisioning by exposing actual usage alongside requested resources in the Explorer.

By adding Kubernetes CPU and memory metrics to the Explorer, CloudZero enables you to:

• View cost and resource usage grouped by cluster, namespace, and workload.
• Compare requested CPU and memory resources to actual usage.
• Identify over-provisioned workloads through efficiency scores.
• Drill down into specific workloads to analyze resource usage in more detail.

Prerequisite

This feature is only available if you are using the CloudZero Agent for Kubernetes. There is no minimum version requirement.

Viewing Kubernetes Usage Metrics

Once the Kubernetes Usage Metrics feature has been enabled for your organization, the data appears automatically in the Explorer. No additional setup or manual tagging is required.

Access the Data

1. Log in to CloudZero and navigate to the Explorer.
2. In the Group By drop-down menu, select the Cluster, Namespace, or Workload Kubernetes Dimension.
3. View the Kubernetes Usage Data column.

Note: Filtering is limited to cluster, namespace, and workload dimensions. Applying filters on unsupported dimensions will result in the corresponding Usage column being omitted.

Example Kubernetes usage metrics grouped by cluster:

Grouped by namespace:

Grouped by workload:

Available Metrics

The Kubernetes Usage Data (Used / Requested) column displays the following metrics for each Kubernetes entity:

Efficiency scores are color-coded based on these thresholds:

• Green: ≥ 80%
• Yellow: 50–79%
• Red: < 50%

Note: Efficiency is capped at 200 percent for display purposes.

The following is an example of Kubernetes metrics color-coded by efficiency score:

Select any cluster, namespace, or workload row to drill into more details, including associated costs and resource profiles.

How Metrics Are Calculated

CloudZero collects pod-level data hourly and smooths metrics to remove short-term spikes and accurately reflect stable workload behavior.

There are three categories of metrics:

• Actual usage
• Requested resources
• Efficiency score

Usage

Usage represents the actual CPU or memory consumed by pods. To prevent brief spikes from skewing the data, CloudZero:

1. Records average usage per pod each hour.
2. Removes the top 5% of spiky hours per pod.
3. Averages remaining values per pod.
4. Sums usage across all pods in the grouping, such as cluster, namespace, or workload.

Requests

Requests represents the CPU or memory requested by pod configuration. To reflect typical configuration rather than point-in-time settings, CloudZero:

1. Captures requested resource value per pod each hour.
2. Averages over time per pod.
3. Sums requests across all pods in the grouping, such as cluster, namespace, or workload.

Efficiency

Efficiency measures how effectively workloads use requested resources. To identify over- or under-provisioning while smoothing out irregular spikes, CloudZero:

1. Efficiency is calculated as usage divided by requests per pod per hour.
2. The top 5 percent of noisy hours are removed per pod.
3. Remaining values are averaged per pod and then across the workload.
4. The final efficiency value is capped at 200 percent for display.

Note that efficiency scores are designed specifically to highlight how much less you're using than you've requested. If you use more than you request, the score will still be color-coded green because over-usage is evaluated separately and is not reflected in this metric.

Efficiency Capping

Efficiency Capping: Efficiency values are capped at 200 percent after the final workload level calculation. Capping does not affect filtering or averaging logic. It is applied only to keep displayed values easy to interpret in the Explorer.