Docs

Volcano Engine Function as a Service

Resource management

Reserved instance

This topic describes features related to reserved instances.

Instance types

Based on the instance usage mode, function instances fall into two types: elastic instances and reserved instances.

Event functions, web applications, and tasks

Type	Elastic instance	Reserved instance
Characteristics	Instance allocation and release are fully managed by the veFaaS system. veFaaS automatically scales instances based on the number of invocations. Instances are created when invocations increase and destroyed when invocations drop. If an instance does not process requests for a period of time, it is automatically destroyed. The first invocation incurs a delay in the instance cold start.	Instance allocation and release are fully managed by the user. After reserved instances are started, they remain running until the user actively releases them. In actual business processing, veFaaS preferentially routes function invocation requests to reserved instances. When concurrent requests exceed the capacity of reserved instances, on-demand elastic instances handle the additional requests.
Application scenarios	Regular business scenarios.	By pre-initializing instances, reserved instances eliminate the latency caused by cold starts and on-demand scaling. They are especially suitable for high-traffic, low-latency scenarios such as e-commerce promotions, live streaming events, and Red Envelope campaigns.
Billing	Elastic instances are billed from the start of function execution to the end of execution. Fees are incurred only when functions are invoked. If no invocation requests are initiated, no instances are allocated, and no fees are incurred. For specific pricing and billing, see Product billing.	Reserved instances are billed after they are created until you actively release them. You are charged for reserved instances as long as they are not released, even if they do not execute any requests. This is because you are responsible for applying for and releasing reserved instances. For specific pricing and billing, see Product billing.

Microservices applications

Microservices applications also support both reserved instances and elastic instances. Reserved instances are always running, while elastic instances are created on demand. Their differences are as follows:

The number of reserved instances for a microservices application must be at least 1, which means at least one instance must always be running.

Users define the autoscaling policy for microservices apps, and veFaaS scales instances based on the user-defined policy.

For microservices applications, both reserved and elastic instances are billed at the reserved instance price.

Configure static reserved instances

Configuring static reserved instances effectively prevents request latency caused by cold starts of elastic instances during peaks of function invocation.

tip

When you configure static reserved instances for the first time, go to Release > Minimum instances.

In the top navigation bar, select the region where your business is located.

In the function list, find the target function and click its name to go to its details page.

Click the Resource management tab. In the Maximum resources section, edit Instance count range to modify the minimum number of instances.

veFaaS will maintain at least this number of continuously running instances until the application stops.

Configure the scheduled scaling policy for instance numbers

When your function exhibits clear periodic patterns or predictable traffic spikes, you can configure scheduled policies based on business characteristics to dynamically scale instances. This improves instance utilization and reduces resource waste.

In the top navigation bar, select the region where your business is located.

In the function list, find the target function and click its name to go to its details page.

Click the Resource management tab. Click Create in the Scheduled policy section.

In the pop-up panel, click Add scheduled policy to configure a scheduled scaling policy.

You can add multiple scheduled policies. Each function can have up to 10 policies.

Parameter	Description
Policy name	Specify a custom name for the scheduled scaling policy. The scheduled policy name must be unique and must be 1 to 64 characters in length. The name can contain letters, digits, hyphens (-), and underscores (_), and cannot start with a digit, hyphen (-), or underscore (_).
Reserved instances	Specify the number of reserved instances for the function when this policy is triggered. The specified number of reserved instances cannot exceed the instance range, such as [0,10].
Policy cycle	Select the effective cycle of the scheduled policy. Daily: The scheduled policy takes effect during the specified time period each day. Weekly: Select the days of the week (natural week) on which the policy takes effect. The policy takes effect during the specified time period on the selected days. Monthly: Select the dates of the month (natural month) on which the policy takes effect. The policy takes effect during the specified time period on the selected dates.
Start time	Set the start time of the policy from the drop-down list.
Duration	Set the duration for which the scheduled policy remains in effect after it is triggered. Unit: minutes.
Policy validity period	Set the validity period of the policy. We recommend setting different validity periods for different scheduled policies.

Check the policy effective time and the preview chart of reserved instance count, and then click Save.

(Optional) View the effective status of the scheduled scaling policy.

After configuration, the scheduled policy is enabled by default. You can edit, delete, and disable it subsequently.

When the scheduled policy takes effect, instances are scaled based on the policy, and the generated reserved instances have a reserved identifier. For more information, see Resource function statistics.

View the number of reserved instances

You can view the number of reserved instances for all functions in the current region on the Resource allocation page.

In the top navigation bar, select the region where your business is located.

In the left navigation pane, click Function > Resource allocation.

In the Allocation details section, view the value in the Instance count range column for the target function. The minimum value of the range is the current number of reserved instances.

Increase the memory quota for reserved instances

If your function has insufficient memory, you can increase the memory by adding more instances on the Resource management page of the function.

In the top navigation bar, select the region where your business is located.

In the function list, find the target function and click its name to go to its details page.

Click the Resource management tab. Edit Instance count range to modify the maximum number of instances in the Maximum resource section.

(Optional) If the number of instances to be added exceeds the Maximum instances displayed on the page, apply for more quota in the Quota Center.

Select AccountResource for Quota parameter, and select the region where the instance is located.

Click Apply in the Actions column.

Enter the quota limit for the target instance and the reason for the application.

Click Submit to apply for the quota.

Last updated: 2026.08.19 14:06:05