Cost Summary↑
Cost Summary shows where your cluster spend is going and how efficiently your resources are used. You can break the spend down by cluster, namespace, application, deployment, and pod.
View the cost summary↑
Select Day-2 Ops in the CAEPE menu, then scroll to Cost Summary. The card shows:
- Total Namespace Cost: the total cost across your namespaces.
- Efficiency: how much of the requested CPU and memory is actually used.
- Cost Allocation: a chart of cost by namespace for the last 7 days.
View the cost breakdown↑
To see more detail, select View Information on the Cost Summary card. The Cost Breakdown page opens.
-
Select how to group the costs:
- Cluster
- Namespace
- Application
- Deployment
- Pod
-
Select a time range: Last 7 days, Last 1 month, Last 3 months, or Specific date/range.
The page then shows:
- Total Cost and Efficiency for the selected grouping and time range.
- Cost Allocation: a chart of cost for each item. Use the buttons above the chart to switch between a bar chart and a line chart.
- Cost Breakdown: a table with the cost of each item, split into the following columns:
| Column | Description |
|---|---|
| Name | The cluster, namespace, application, deployment, or pod. |
| CPU | Cost of processing power (CPU). |
| GPU | Cost of graphics processors (GPU). |
| MEMORY | Cost of memory. |
| PV | Cost of persistent volumes (PV), the storage attached to your workloads. |
| Efficiency | CPU and memory used, divided by CPU and memory requested. |
| Total Cost | The total cost for the item. |
The TOTAL row at the bottom of the table adds up the cost of all items.
Note
Cost figures can take a moment to load, because CAEPE collects them from each of your clusters.
Tip
A low efficiency usually means workloads request more CPU or memory than they use. See Workload Optimization for recommended values.
How costs are calculated↑
- CPU, GPU, and memory are charged for whichever is higher: the amount a workload requests or the amount it actually uses. A workload that requests more than it needs is charged for the full request, and a workload that uses more than it requests is charged for what it uses.
- Persistent volumes (PV) are charged for the average size of the storage over the selected time range.
- Efficiency is the amount of CPU and memory used divided by the amount requested. CAEPE combines the CPU and memory figures, giving more weight to whichever costs more. GPU and storage are not included. A workload that uses resources without setting a request is counted as 100% efficient.
For example, a workload that requests 2 CPU cores but uses only 0.5 is charged for 2 cores and has a CPU efficiency of 25%.