This page describes how Compute flexible committed use discounts (flexible CUDs) and legacy Autopilot committed use discounts (CUDs) work with Google Kubernetes Engine (GKE) Autopilot clusters. To learn more about committed use discounts in GKE Standard clusters, refer to the committed use discount pricing through Compute Engine.
Flexible CUDs and the legacy Autopilot CUDs for GKE are spend-based CUDs that provide deeply discounted prices in exchange for committing to use Autopilot workloads for a specified term.
Depending on your usage requirements, you can purchase commitments for Kubernetes Engine (Autopilot Mode) in any of the following ways:
| Flexible CUDs | Legacy Autopilot CUDs |
|---|---|
Use for predictable spend on any of the following products that are eligible
for flexible CUDs:
|
Use for predictable spend on Autopilot resources that aren't eligible for flexible CUDs, like Scale-Out Pods and GPU Pods that don't use the Accelerator class. Not available to purchase on or after October 15, 2024. |
| Applies to any region and any project in a Cloud Billing account. | Applies to all projects within a specific region in a Cloud Billing account. |
|
|
You can purchase flexible and legacy Autopilot commitments from any Cloud Billing account, and the discount applies to any eligible usage in projects paid for by that Cloud Billing account:
- Any usage overage after you exhaust all your commitments (both flexible and legacy Autopilot CUDs) is charged at the on-demand rate.
- When you purchase flexible or legacy Autopilot commitments, you pay the same commitment fee for the entirety of the commitment term, even if the price of applicable usage changes.
- You still receive the same discount percentage on applicable usage in the event of a price change.
- The commitment fee is billed monthly.
For more information about your bill, see Analyze the effectiveness of your spend-based committed use discounts.
Applicable usage for flexible CUDs and legacy Autopilot CUDs
Flexible CUDs and the legacy Autopilot CUDs automatically apply to aggregate GKE Autopilot Pod usage in a region, giving you low, predictable costs, without the need to make any manual changes or updates yourself. This flexibility saves you time and helps you to save more by achieving high utilization rates across your commitments.
Flexible CUDs and the legacy Autopilot CUDs apply to the following:
- All Autopilot Pod workload CPU, memory, and ephemeral storage usage in the region in which you have committed.
- For Performance compute class Pods, legacy Autopilot CUDs apply only to the Autopilot premium. You can separately purchase Compute Engine committed use discounts to cover Compute Engine VMs that run the Performance compute class Pods. Flexible CUDs always apply to Autopilot premium, and also apply to its underlying resources if the resources are eligible to receive Flexible CUDs.
- For Accelerator compute class Pods, CUDs apply only to the Autopilot premium. You can separately purchase Compute Engine resource-based commitments for the Compute Engine virtual machines that run the Accelerator compute class Pods.
Flexible CUDs don't apply to Scale-Out compute class or to GPU Pods that don't use Accelerator class. If you want to use legacy Autopilot CUDs for these use cases, you can purchase the commitments until October 15, 2024.
Flexible CUDs and the legacy Autopilot CUDs don't apply to the cluster management fee, Spot Pods, or to GKE Standard mode compute nodes. GKE Standard mode compute nodes can use committed use discount pricing through Compute Engine.
Order of discount application
Google Cloud first applies and exhausts legacy Autopilot discounts on your CUD commitments, and any eligible usage that isn't covered by the existing legacy Autopilot commitments becomes eligible for new flexible CUDs. If you use any additional resources that take your hourly spend amount beyond your committed hourly spend amount, then the overage usage is not covered by flexible CUDs.
Purchase commitments
You can purchase commitments only at a Cloud Billing account level. For more information about how to purchase a commitment, see Purchasing spend-based commitments.
Before you purchase a commitment, read the Service Specific Terms.
After you purchase a commitment, you can't cancel it. For more information, see Cancelling commitments.
Flexible CUDs
Flexible CUDs add flexibility to your spending capabilities by eliminating the need to restrict your commitments to a single project, or region.
Flexible CUDs pricing
When you purchase flexible commitments for your Cloud Billing account you commit to a minimum amount of hourly spend on Kubernetes Engine (Autopilot Mode) for a 1-year or 3-year term. You do this by committing to spend on resources that are worth a specified minimum amount of on-demand price, every hour, throughout the commitment term. In return for committing to an hourly spend amount, you receive the following discounts:
- 1-year commitment: 28% off on-demand pricing
- 3-year commitment: 46% off on-demand pricing
The commitment that you purchase becomes active within the first hour of its purchase. The discounted hourly spend amount becomes your commitment fee and you are billed this fee monthly. Your commitment fee remains your minimum hourly expenditure throughout the commitment term and you have to pay it even if you don't use resources for which the sum of on-demand prices are equal to your committed hourly spend.
Your commitment fee remains the same even if the on-demand prices for your resources change during your commitment term.
Calculate hourly on-demand commitment for flexible CUDs
You calculate the baseline per hour cost of GKE Autopilot Pod vCPUs + Pod memory across all your GKE Autopilot clusters in the region that you want to benefit from an existing commitment. Any usage beyond that limit will be charged at the regular on-demand price.
For example:
- Assume you are running a total of 97.5 Pod vCPU and 121 GB Pod Memory in
each of two different regions. The Autopilot clusters are located
in regions
us-central1(Iowa) and asia-southeast1 (Singapore), and you're interested in buying a 1-year CUD.
From the pricing table, you can calculate the total hourly commitment cost per region:
Iowa
- 97.5 * $0.0445 per vCPU per hour = ~$4.339 per hour
- 121 * $0.0049225 per GB per hour = ~$0.596 per hour
- For a total of $4.935 per hour in committed use hourly pricing.
Singapore
- 97.5 * $0.0549 per vCPU per hour = ~$5.353 per hour
- 121 * $0.0060729 per GB per hour = ~$0.735 per hour
- For a total of $6.088 per hour in committed use hourly pricing.
Calculating the monthly cost (720 hours in a month):
- Iowa
- On-demand pricing = $4.935 per hour * 720 hours = ~$3,553.20 per month
- After 28% 1-year CUD discount = ~$2,558.30 per month
- Total savings of $994.9 per month.
- Singapore
- On-demand pricing = $6.088 per hour * 720 hours = ~$4,383.36 per month
- After 28% 1-year CUD discount = ~$3,156.02 per month
- Total savings of ~$1227.16 per month.
Once you make the commitment, you're charged that amount even if you decide to stop or scale down the actual number of vCPUs or RAM during the month.
Legacy Autopilot CUDs
The following sections cover pricing, calculating hourly on-demand commitment and recommendations for the legacy Autopilot CUDs.
Legacy Autopilot CUD pricing
Legacy Autopilot CUDs give you a 20% off on-demand pricing for a one-year commitment and a 45% off on-demand pricing for a three-year commitment. These discount percentages are the same in every region.
See committed use discount pricing for pricing details.
Key points to keep in mind for the legacy Autopilot CUDs:
- Only apply to CPU, memory, and ephemeral storage of Autopilot pods.
- Do not apply to cluster management fee, or to GKE Standard mode compute nodes.
- Apply to all Autopilot workloads in a given region.
- Measured in dollars per hourly on-demand commitment.
Calculate hourly on-demand commitment for legacy CUDs
You calculate the baseline per hour cost of GKE Autopilot Pod vCPUs + Pod memory across all your GKE Autopilot clusters in the region that you want to benefit from an existing CUD.
For example:
- Assume you are running a total of 97.5 Pod vCPU and 121 GB Pod Memory in each of two different regions. The Autopilot clusters are located in regions us-central1 (Iowa) and asia-southeast1 (Singapore), and you're interested in buying a 1-year CUD.
From the pricing table, we can calculate the total hourly commitment cost per region:
Iowa
- 97.5 * $0.0445 per vCPU per hour = ~$4.339 per hour
- 121 * $0.0049225 per GB per hour = ~$0.596 per hour
- For a total of $4.935 per hour in committed use hourly pricing.
Singapore
- 97.5 * $0.0549 per vCPU per hour = ~$5.353 per hour
- 121 * $0.0060729 per GB per hour = ~$0.735 per hour
- For a total of $6.088 per hour in committed use hourly pricing.
Calculating the monthly cost (720 hours in a month):
- Iowa
- On-demand pricing = $4.935 per hour * 720 hours = ~$3,553.20 per month
- After 20% 1-year CUD discount = ~$2,842.56 per month
- Total savings of ~$710.64 per month.
- Singapore
- On-demand pricing = $6.088 per hour * 720 hours = ~$4,383.36 per month
- After 20% 1-year CUD discount = ~$3506.69 per month
- Total savings of ~$876.67 per month.
Once you make the commitment, you're charged that amount even if you decide to stop or scale down the actual number of vCPUs or RAM during the month.