You can set a resource reservation to establish a capacity baseline for the Load Balancer Capacity Units (LCUs) of your Application Load Balancer (ALB) instance. During high-load scenarios, ALB can instantly use this reserved capacity to manage traffic peaks. This eliminates the need to wait for auto-scaling and improves the availability and stability of your service. The ALB instance still scales automatically, but its performance will not drop below the reserved capacity. You are billed hourly for the number of reserved LCUs.
Resource reservation is ideal for the following scenarios:
You are launching an operational activity that will cause a sudden traffic spike. You want to ensure that your ALB instance can handle the peak traffic during the event.
Your business is burstable and you cannot effectively predict traffic spikes.
A service that is being published or migrated requires the ALB instance to have high performance from the start, rather than waiting for auto-scaling.
You need to maintain a deterministic capacity to meet your service requirements.
You are performing a migration between Server Load Balancer instances and want the target instance's performance to match the source instance.
Applicability
The resource reservation feature is not enabled by default. To use this feature, contact your account manager to request access.
Resource reservation is supported only on upgraded ALB instances. You can manually migrate services from a non-upgraded ALB instance to an upgraded one using the ALB instance cloning feature.
Only Standard and WAF-enabled ALB instances support resource reservation. Basic ALB instances do not. You can upgrade the edition of an ALB instance.
Configure a resource reservation
To configure a resource reservation, the instance must have at least one listener in the Running state.
An Internet-facing ALB instance provides Internet access through elastic IP addresses (EIPs) by default. Resource reservation does not include bandwidth reservation for EIPs. To increase your bandwidth, purchase and add your EIPs to an Internet Shared Bandwidth instance. If you use an Anycast EIP, increase the peak bandwidth of the Anycast EIP.
The reserved LCU capacity is evenly distributed among the zones where the ALB instance is deployed. To achieve optimal performance for an ALB instance with resource reservation, deploy the instance in at least two zones, configure the same number of backend services in each zone, and enable cross-zone load balancing.
Console
In the ALB console, click the ID of the target instance. On the Instance Details page, select the Resource Reservation tab and click LCU Resource Reservation.
Select a method for Reserved Capacity Estimation:
Estimate Based on Reference Value: If you have historical traffic data, select the target ALB instance from the Reference SLB Instance drop-down list and check the historical Peak LCU value in the chart. The Peak LCU value indicates the maximum processing performance that the ALB instance used during historical traffic peaks.
Manual Estimation: If you do not have historical traffic data, enter the Peak Bandwidth and New Connections based on your traffic estimates for the ALB instance, and then check the system-calculated Estimated LCU.
Enter the Total Reserved LCUs based on the reference value and click OK. The minimum resource reservation is 100 LCUs. The maximum capacity is limited by the quota.
For latency-sensitive scenarios, such as financial transactions, you can deploy resources in one of your deployment sets with a low-latency network policy to reduce forwarding latency. To use this feature, contact your account manager to request access.
API
Call ModifyCapacityReservation to configure a resource reservation.
View a resource reservation
Console
In the ALB console, click the ID of the target instance. On the Instance Details page, select the Resource Reservation tab.
In the Resource Reservation area, view information such as Reservation Status and Reserved LCU.
In the Load Balancer LCU Usage area, you can compare the Reserved LCU and Peak LCU values to monitor the usage of your Reserved LCU.
For more information, see Metrics for resource reservation.
API
Call DescribeCapacityReservation to query the status of a resource reservation.
The following table describes the possible status values of a resource reservation.
Status | Status descriptions | Can be modified or canceled |
Pending | The resource reservation is being configured. | No |
Ready | The reserved capacity is ready and available for use. | Yes |
Failed | The resource reservation request cannot be completed at the moment. | Yes |
Rebalancing | A zone has been added or removed, and the Server Load Balancer instance is rebalancing its capacity. Any change in the number of zones where the ALB instance is deployed triggers automatic rebalancing to evenly redistribute the total capacity across the zones. | No |
Modify or cancel a resource reservation
You can upgrade the reserved capacity for each ALB instance an unlimited number of times. However, you can downgrade or cancel the reservation a maximum of two times per day in total.
If the underlying performance metrics for your reserved capacity exceed the maximum auto-scaling performance of the ALB instance, you cannot downgrade or cancel the reservation for seven days after it is successfully created.
Console
In the ALB console, click the ID of the target instance. On the Instance Details page, select the Resource Reservation tab.
To modify the reserved LCU capacity, click LCU Resource Reservation, set a new value for Total Reserved LCUs, and click OK.
To cancel the resource reservation, click Cancel Resource Reservation and then click OK.
API
Call ModifyCapacityReservation to modify or cancel a resource reservation.
Billing details
ALB LCU fees are charged hourly. The billing cycle is one hour, and usage for less than one hour is billed as a full hour. If you modify your reservation within a billing cycle, the system uses the highest number of reserved LCUs during that cycle to calculate the reserved LCU fee.
If the actual LCU usage is less than or equal to the number of reserved LCUs, the bill includes only the reserved LCU fee:
Reserved LCU fee = Unit price per reserved LCU × Number of reserved LCUs
If the actual LCU usage is greater than the number of reserved LCUs, the bill includes both the LCU fee and the reserved LCU fee:
LCU fee = Unit price per LCU × (Actual LCU usage - Number of reserved LCUs)Reserved LCU fee = Unit price per reserved LCU × Number of reserved LCUs
The unit price of a reserved LCU is the same as the unit price of an LCU. The actual price is subject to the price on the purchase page.
Resource plans cannot be used to offset fees for reserved LCU capacity. A resource plan can offset fees only for the portion of LCU usage that exceeds the reserved capacity.
The following table shows the billing details for an ALB instance with different resource usage and reservation configurations in each billing cycle.
Billing cycle | Key operations and events | Actual LCU usage | Number of reserved LCUs | Fee |
10:00:00 to 10:59:59 |
| 20 | - | LCU fee for this hour = USD 0.007/unit × 20 = USD 0.14 |
11:00:00 to 11:59:59 |
| 30 | 100 | Reserved LCU fee for this hour = USD 0.007/unit × 100 = USD 0.7 |
12:00:00 to 12:59:59 |
| 150 | 100 | LCU fee for this hour = USD 0.007/unit × (150 - 100) = USD 0.35 Reserved LCU fee for this hour = USD 0.007/unit × 100 = USD 0.7 |
13:00:00 to 13:59:59 |
| 110 | 120 | Reserved LCU fee for this hour = USD 0.007/unit × 120 = USD 0.84 |
14:00:00 to 14:59:59 |
| 30 | 120 | Reserved LCU fee for this hour = USD 0.007/unit × 120 = USD 0.84 |
Quotas
To adjust a quota, contact your account manager to request an increase.
Quota name | Description | Default value |
alb_quota_reserved_capacity_units_per_loadbalancer | The maximum reserved LCU capacity for each ALB instance | 5,000 |
alb_quota_reserved_capacity_units_per_region | The maximum reserved LCU capacity per region | 20,000 |
FAQ
How do I determine the LCU capacity to reserve? What are the consequences of reserving too much or too little?
Reserving too much capacity results in unnecessary costs because you are billed for reserved LCUs even if they are not used. Reserving too little capacity may not fully cover traffic peaks. This means some traffic will still have to wait for auto-scaling, which can affect service stability. We recommend that you estimate your resource needs as described in the Configure a resource reservation section. Use this estimation along with stress testing results to determine a reasonable value. You can adjust the reservation later based on your observed LCU usage.
After I downgrade or cancel a resource reservation, when does the billing change?
A downgrade or cancellation takes effect at the beginning of the next billing cycle. For example, if you successfully cancel a resource reservation at 14:30, you are still billed for the maximum number of reserved LCUs that were active during the 14:00:00 to 14:59:59 billing cycle. Billing for the reservation stops at 15:00.