Use Cost Optimization to analyze and optimize cluster costs - ACK-Expenses and Costs-阿里云帮助中心

Procedure

Log in to the Billing Management console, navigate to the Cost Management > Cost Optimization page, and click Analyze and optimize cluster costs - ACK to open the analysis page for cloud-native scenarios.

Note

The first time you access this feature, an activation page appears. Review the feature's use cases and click Activate Now to activate it for free.

On the optimization page, select your cluster from the drop-down list.

The tabs below display the cost analysis for the selected cluster.

Cluster cost overview

The following table describes the features on the cluster cost visualization dashboard.

Feature	No.	Description
Assess cluster waste based on cost trends	①	The total cost. If the percentage change from the previous day is blue, the cost has decreased. If the percentage change from the previous day is red, the cost has increased.
	②	The yellow line shows cost, and the blue line shows the actual cluster capacity. Normally, these two lines are correlated. If the lines move in different directions, the cost per core may be unusual. Check if more expensive resources are being used.
	③	The cost trend over a period of time.
Allocate departmental costs by namespace	④	A cluster can contain nodes with different instance types and billing methods. When you allocate costs by namespace, you cannot rely only on the resource requests of each namespace. You must also consider the nodes where the Pods in the namespace are running. Cost analysis converts the real-time cost of each node. The cost for a namespace is then calculated as follows: `Σ (Pod resource request / node capacity) * node unit price` This method provides an accurate estimate of namespace costs. However, the estimated costs may differ from the actual cluster bill because of discounts, coupons, or subscriptions. To allocate costs at the namespace level, multiply the namespace's cost percentage by the total cluster cost.
Analyze cloud service costs by trend and proportion	⑤	A cluster uses various cloud services. Different usage patterns and billing models can lead to large cost differences. You can view the cost trends and breakdown of different cloud services to make informed cost decisions.

Cluster application waste analysis

On the Cluster Application Waste Analysis page, you can filter and view resource and cost information by namespace. A namespace often corresponds to a department or team in your company.

Feature	No.	Description
Assess namespace resource waste by using CPU and memory metrics	①	Displays the CPU Resource Usage and Memory Resource Usage charts.
	②	Shows trends for resource allocation versus actual consumption. The blue line represents allocated resources, and the red line represents consumed resources. When a Pod is scheduled, the node allocates resources to it. However, the resources a container actually uses often differ from this allocation. This chart shows this relationship, helping you reduce resource waste.
	③	The trend of resource utilization over a period of time.
	④	Pod CPU utilization and Pod memory utilization are the ratios of consumed resources to allocated resources. The amount of allocated but unused resources indicates waste for a Pod. By finding the most wasteful Pods in a namespace, you can identify the primary applications that cause resource waste. This helps you analyze the cause and create targeted optimization strategies.
Analyze departmental costs	⑤	Due to discounts, coupons, and subscriptions, the estimated namespace costs may differ from the actual cluster bill. You can achieve namespace-level cost allocation by multiplying the namespace's cost percentage by the total cost from the cluster's bill.

Cluster node pool cost analysis

The node pool dashboard provides cost insights from a resource-centric view, which helps you analyze node pool resource usage and select billing methods. This view allows you to plan and govern costs based on different node pools. For example, resources such as GPU node pools may be shared across multiple departments, making it hard to analyze costs by namespace. The node pool view allows you to set policies and optimize costs directly at the resource level.

Feature	No.	Description
Assess node pool waste based on cost trends	①	The total cost. If the percentage change from the previous day is blue, the cost has decreased. If the percentage change from the previous day is red, the cost has increased.
	②	The cost trend over a period of time.
	③	The trend of the node pool's cost share over a period of time.
Predict potential savings with different billing methods	④	Statistics on the billing methods of nodes in the node pool.
Predict potential savings with different billing methods	⑤	This example assumes all nodes in the node pool are pay-as-you-go. Shows the predicted hourly savings and percentage if you switch all nodes to subscription. Shows the predicted hourly savings and percentage if you switch all nodes to preemptible instances. By comparing the costs of different billing methods and reviewing these savings predictions, you can optimize your resource mix and billing strategy.

Cluster application cost analysis

The application dashboard allows you to use label wildcards to filter for applications and view their cost and resource statistics. This view is ideal for scenario-based cost optimization, such as for big data, AI, or elastic workloads.

By matching label wildcards, you can monitor the cost and resource usage of a single application or a group of related applications. For example, in a big data workflow, you can apply a consistent label to all applications in the process to analyze the costs for the entire workflow.

Feature	No.	Description
Assess business costs from an application perspective	①	Shows metrics such as application cost, current replica count, application runtime, total core-hour consumption, and the application's resource utilization within the cluster or namespace.
	②	The trend of application costs over a period of time.
	③	The trend of the application's Pod scale over a period of time.
	④	The trend of the application's price per core-hour over a period of time.
Analyze business costs from a Pod perspective	⑤	Shows metrics such as a Pod's CPU and memory utilization, price per core-hour, cost, total core-hour consumption, and runtime.

FAQ

Q1: Why is no data displayed after I enable cost analysis?

A: Check that you have granted the required permissions to the RAM role. For detailed instructions, see Step 5 of the procedure for enabling cost analysis. Cost analysis requires monitoring data and cost data from the cluster. Monitoring data collection begins within 3 minutes of enabling the feature. Cost data becomes available at 08:00 the day after you enable the feature.

Q2: Why does the sum of namespace costs not match my actual bill?

A: Namespace costs are estimates based on list prices and do not reflect your actual bill. Therefore, differences can occur when your cluster's actual spending includes deductions from coupons, discounts, or savings plans. However, you can still allocate costs by multiplying each namespace's cost percentage by the total cluster cost.

Q3: Why is cost data unavailable for Pods on certain nodes?

A: Check that you have granted the required permissions to the RAM role. Ensure that the authorization policy includes the "ecs:DescribeDisks" permission. For detailed instructions, see Step 5 of the procedure for enabling cost analysis.

Q4: Why don't the bills in the analysis show all cloud services that my cluster uses?

A: Cost analysis includes only cloud services used exclusively by the current cluster. Shared cloud services are not included.