A CU capacity reservation locks in a minimum processing capacity for a NAT Gateway so that the gateway delivers predictable performance during a traffic spike. By default, NAT Gateway scales automatically based on actual CU usage, and scaling takes minutes when traffic surges, which may cause brief packet loss.
Capacity reservation is available by whitelist. To use it, contact your account manager to request access.
Use cases
Configure a capacity reservation for a NAT Gateway in the following cases:
Traffic spike guarantee: Locks in a minimum resource capacity for workloads with sustained peak traffic, which avoids packet loss caused by scaling delays.
Event capacity reservation: Locks in resources before an expected traffic peak and releases them after the event, which balances stability and cost.
Business launch testing: Locks in a baseline resource capacity for a new service, so you can evaluate the elastic policy after traffic data is collected.
How it works
When you configure resource reservation, you can use the CU estimator to estimate the required CU reservation by entering your bandwidth, new connections per second (CPS), concurrent connections (CC), and packets per second (PPS).
Actual traffic is lower than the reserved capacity: Performance is guaranteed and billed based on the reserved CUs.
Actual traffic exceeds the reserved capacity: NAT Gateway scales out automatically.
A capacity reservation takes effect after the reserved resources are ready. The system returns the performance metrics of the reserved CUs, including bandwidth, new connections per second (CPS), concurrent connections (CC), and data packets processed per second (PPS). Compare these metrics with your business needs to check whether the reserved capacity is sufficient.
Limits and quotas
The following limits apply to NAT Gateway capacity reservations:
CU limits: The minimum value is 9,000 CUs (about 20 Gbps), and the default upper limit is 11,250 CUs (about 25 Gbps). The upper limit is adjustable.
Scale-in limit: Two times per day, including canceling a capacity reservation. After using up the daily quota, you must wait until the next day for it to reset.
Configure a capacity reservation
Console
Log on to the NAT Gateway console and click the ID of the target instance to go to the details page of the instance.
On the Resource Reservation tab, click Configure CU Resource Reservation.
On the configuration page, use the LCU Calculator to estimate the number of CUs to reserve. Enter the following metrics:
Bandwidth Peak (Bidirectional Sum): the maximum bandwidth of your service. Unit: Gbps.
New Connection Rate: the peak number of new connections per second.
Concurrent Connections: the maximum number of connections maintained at the same time.
Packet Processing Rate: the peak number of data packets processed per second.
In Total Reserved CUs, enter the total number of CUs to reserve based on the estimated value and your actual business needs. Select the checkbox to acknowledge the fees.
Submit the capacity reservation request. Last Reservation Status changes to Preparing. After the resources are ready, the status changes to Reserved.
Modify or cancel a capacity reservation
Scale out: No limit. You can increase the reserved CU count at any time.
Scale in: Twice per day, including cancellations. Once exhausted, you must wait until the next day.
Console
Modify a capacity reservation
You can modify or cancel a capacity reservation when the reservation status is Reserved or Reservation Failed.
On the Resource Reservation tab of the product page of the instance, click Configure CU Resource Reservation.
Enter the new number of CUs to reserve and click OK.
Cancel a capacity reservation
After you cancel a capacity reservation, NAT Gateway releases the reserved resources and returns to full automatic scaling mode. The cancellation counts toward the scale-in quota of the day.
On the Resource Reservation tab of the details page of the instance, click Cancel Resource Reservation.
In the confirmation dialog box, click OK to cancel the reserved capacity.
Billing
The reserved CU fee of a NAT Gateway is charged hourly, and the billing cycle is 1 hour. Within a billing cycle, the system compares the initial number of reserved CUs with the number of reserved CUs that take effect after a configuration change, and calculates the reserved CU fee based on the maximum value. If the usage duration within a billing cycle is less than 1 hour, the fee is prorated based on the actual usage duration.
When the number of CUs actually consumed is less than or equal to the number of reserved CUs, the bill contains only the reserved CU fee:
Reserved CU fee = Reserved CU unit price × Number of reserved CUs
When the number of CUs actually consumed is greater than the number of reserved CUs, the bill contains the reserved CU fee and the excess CU fee:
Reserved CU fee = Reserved CU unit price × Number of reserved CUsExcess CU fee = CU unit price × (Number of CUs actually consumed - Number of reserved CUs)
The reserved CU unit price is the same as the Standard CU unit price. The actual purchase price is subject to the buy page.
The excess portion is calculated by minute. Only the minutes in which actual usage exceeds the reserved capacity incur excess fees. Minutes that do not exceed the reserved capacity are not billed.
Reserved CU capacity cannot be offset by resource plans. The excess portion can be offset by a resource plan only when the number of CUs actually consumed exceeds the number of reserved CUs.
Billing examples
The following table shows the fees of a cross-zone disaster recovery NAT Gateway in the China (Hangzhou) region in each billing cycle. The fees vary with the resource usage and the capacity reservation configurations. In the following examples, the CU unit price is USD 0.034/CU.
Billing cycle | Key operations | Exceeded CUs | Reserved CUs | Fees |
10:00:00-10:59:59 | Routine operation without a capacity reservation. 200 CUs are actually consumed. | 200 No reservation. Billed for actual CU usage. | - | The CU fee for this hour = USD 0.034/CU × 200 = USD 6.80 |
11:00:00-11:59:59 | A capacity reservation of 1,000 CUs is configured for an event and becomes ready at 11:45:00. 300 CUs are actually consumed, lower than reserved CUs. | 0 | 1,000 | The reserved CU fee for this hour = USD 0.034/CU × 1,000 = USD 34.00 |
12:00:00-12:59:59 | The initial capacity reservation is 1,000 CUs. During peak traffic, consumption exceeds the reservation every minute, totaling 500 CU in excess. | 500 | 1,000 | The excess CU fee for this hour = USD 0.034/CU × 500 = USD 17.00 The reserved CU fee for this hour = USD 0.034/CU × 1,000 = USD 34.00 |
13:00:00-13:59:59 | The initial capacity reservation is 1,000 CUs. Reservation increased to 1200 CU, effective 13:05. Actual consumption is 1100 CU, never exceeding the reservation. | 0 | 1,200 | The reserved CU fee for this hour = USD 0.034/CU × 1,200 = USD 40.80 |
14:00:00-14:59:59 | The initial capacity reservation is 1,200 CUs. Event ended, reservation canceled effective 14:20. Actual consumption is 200 CU, never exceeding the reservation. | 0 | 1,200 | The reserved CU fee for this hour = USD 0.034/CU × 1,200 = USD 40.80 |
Quota
The following quota applies to NAT Gateway capacity reservations:
Quota name | Description | Default value | Adjustable |
| The maximum number of reserved CUs that can be configured for a single NAT Gateway instance | 11,250 | Yes. The value can be increased to a maximum of 45,000. Go to the Quota Management page to apply for an increase. |
FAQ
What is the relationship between capacity reservation and automatic scaling?
Capacity reservation provides a minimum performance baseline for your NAT gateway. Without it, the NAT gateway relies entirely on auto-scaling to scale on demand. With capacity reservation, the NAT gateway delivers predictable performance up to the reserved capacity. Any traffic beyond that continues to be handled by auto-scaling.
What happens if I reserve too many or too few CUs?
Reserving too many CUs: You pay for unused capacity. When actual consumption falls below the reservation, you are still billed for the full reserved amount.
Reserving too few CUs: Traffic peaks still trigger auto-scaling, which introduces latency and may cause brief packet loss, impacting service stability.
Estimate the number of CUs to reserve using the CU calculator, and then determine the initial value based on your stress test results.
When does billing change after I scale in or cancel a capacity reservation?
Changes take effect at the start of the next billing cycle. For example, if you cancel a reservation at 14:30, the 14:00–14:59 cycle is still billed based on the highest reserved CU count effective during that cycle. No reserved CU fees are incurred from 15:00 onward.