Quotas in Amazon Cognito - Amazon Cognito

Quotas in Amazon Cognito

Amazon Cognito has default quotas, formerly referred to as limits, for the maximum number of operations that you can perform in your account. Amazon Cognito also has quotas for the maximum number and size of Amazon Cognito resources.

Each Amazon Cognito quota represents a maximum volume of requests in one AWS Region in one AWS account. For example, your apps can make API requests at up to the Default quota (RPS) rate for UserAuthentication operations against all of your user pools in US East (N. Virginia). Your apps in Asia Pacific (Tokyo) can produce the same volume of requests against all of your user pools in their own Region. AWS can only grant a quota increase request in one Region at a time. A successful quota increase in US East (N. Virginia) has no effect on your maximum request rate in Asia Pacific (Tokyo).

Understanding API request rate quotas

Quota categorization

Amazon Cognito enforces a maximum request rate for API operations. For more information about the API operations that Amazon Cognito makes available, see Amazon Cognito API and endpoint references For user pools, these operations are grouped into categories of common use cases like UserAuthentication or UserCreation. For a list of user pool API operations by category, see Amazon Cognito user pools API operation categories and request rate quotas.

In the Service Quotas console, you can track your quota usage by category user pools and identity pools.If the request rate of your Amazon Cognito user pools or exceeds a quota, you can purchase additional capacity. You can track your user pool quota usage by category and purchase quota increases in the Service Quotas console.

Operation quotas are defined as the maximum number of requests per second (RPS) for all operations within a category. The Amazon Cognito user pools service applies quotas to all operations in each category. For example, the category UserCreation includes four operations: SignUp, ConfirmSignUp, AdminCreateUser, and AdminConfirmSignUp. It's allocated with a combined quota of 50 RPS. If multiple operations take place at the same time, each operation within this category can call up to 50 RPS separately or combined.

Note

Category quotas only apply to user pools. Amazon Cognito applies each identity pool quota to a single operation. For both per-category and per-operation request rate quotas, AWS measures the aggregate rate of all requests from all user pools or identity pools in your AWS account in one Region.

Amazon Cognito user pools API operations with special request rate handling

Operation quotas are measured and enforced for the combined total requests at the category level, except for the AdminRespondToAuthChallenge and RespondToAuthChallenge operations, where special handling rules are applied.

The UserAuthentication category includes four operations in the Amazon Cognito user pools API: AdminInitiateAuth, InitiateAuth, AdminRespondToAuthChallenge, and RespondToAuthChallenge. Additionally, user authentication in the hosted UI contributes to this quota. The InitiateAuth and AdminInitiateAuth operations are measured and enforced per category quota. The matching operations RespondToAuthChallenge and AdminRespondToAuthChallenge are subject to a separate quota that is three times the UserAuthentication category limit. This elevated quota accommodates multiple authentication challenges set up in your apps. The quota is sufficient to cover the large majority of use cases. After your app makes up to three responses to authentication challenges, additional requests count toward the UserAuthentication category quota. Multi-factor authentication (MFA), device authentication, and custom authentication are all examples of challenge prompts that you might engineer into your user pool.

For example, if your quota for the UserAuthentication category is 80 RPS, you can call RespondToAuthChallenge or AdminRespondToAuthChallenge at a rate up to 240 RPS (3 * 80 RPS). If your user pool prompts for four rounds of challenge per authentication and 70 users sign in per second, then the total RespondToAuthChallenge is 280 RPS (70 x 4), which is 40 RPS above the quota. The extra 40 RPS is added to 70 InitiateAuth calls, making the total usage of UserAuthentication category 110 RPS (40 + 70). Because this value exceeds the category quota set at 80 RPS by 30 RPS, Amazon Cognito throttles requests from your app.

Monthly active users

When Amazon Cognito calculates user pool billing, it charges you a rate for each monthly active user (MAU). Consider your current and projected MAU count in your planning for quota increase requests. A user is counted as a MAU if, within a calendar month, there is an identity operation related to that user. The activities that make a user active include the following.

  • Sign-up or administrative creation of a user

  • Sign-in

  • User account confirmation or attribute verification

  • Password reset

  • Change user attributes, group membership, or MFA preferences

  • Query detailed attributes of a user

  • User activation, deactivation or deletion

Note

The category Query detailed attributes of a user includes the API operation AdminGetUser, but not ListUsers. A detailed user-by-user query in a large user pool can have a significant impact on your AWS bill. To avoid excess charges, collect user data with ListUsers or store user information in an external database.

Managing API request rate quotas

Identify quota requirements

Important

If you increase Amazon Cognito quotas for categories such as UserAuthentication, UserCreation, or AccountRecovery, you may need to increase quotas for other AWS services. For example, messages that Amazon Cognito sends with Amazon Simple Notification Service (Amazon SNS) or Amazon Simple Email Service (Amazon SES) can fail if request rate quotas are insufficient in those services.

To calculate quota requirements, determine how many active users will interact with your application in a specific time period. For example, if you expect your application to sign in an average of one million active users within an eight-hour period, then you must be able to authenticate an average of 35 users per second.

In addition, if you assume that the average user session is two hours, and you configure tokens to expire after an hour, each user must refresh their tokens once during their session. The required average quota for the UserAuthentication category to support this load is 70 RPS.

If you assume a peak-to-average ratio of 3:1 by accounting for the variance of user sign-in frequency during the eight-hour period, then you need the desired UserAuthentication quota of 200 RPS.

Note

If you call multiple operations for each user action, you must sum up the individual operation call rates at the category level.

Optimize request rates for quota limits

Because increasing API rate limits adds costs to your AWS bill, consider adjustments to your usage model before you request a quota increase. The following are some examples of app architecture that optimizes request rates.

Retry the attempt after a back-off waiting period

You can catch errors with each API call, and then re-try the attempt after a back-off period. You can adjust the back-off algorithm according to business needs and load. Amazon SDKs have built-in retry logic. For more information, see Tools to Build on AWS.

Use an external database for frequently updated attributes

If your application requires several calls to a user pool to read or write custom attributes, use external storage. You can use your preferred database to store custom attributes or use a cache layer to load a user profile during sign-in. You can reference this profile from the cache when needed, instead of reloading the user profile from a user pool.

Validate JSON web tokens (JWTs) on the client side

Applications must validate JWT tokens before trusting them. You can verify the signature and validity of tokens on the client side without sending API requests to a user pool. After the token is validated, you can trust claims in the token and use the claims instead of making more getUser API calls. For more information, see Verifying a JSON Web Token.

Throttle traffic to your web application with a waiting room

If you expect traffic from a large number of users signing in during a time-bound event, such as taking an exam or attending a live event, you can optimize request traffic with self-throttling mechanisms. You can, for example, set up a waiting room where users can stand by until a session is available, allowing you to process requests when you have available capacity. See the AWS Virtual Waiting Room solution for a reference architecture of a waiting room.

Cache JWTs

Reuse access tokens until they expire. For an example framework with token caching in an API Gateway, see Caching tokens. Instead of generating API requests to query user information, cache ID tokens until they expire, and read user attributes from the cache.

For more information about working with API request rates in AWS, see Managing and monitoring API throttling in your workloads.

Track quota usage

Amazon Cognito generates CallCount and ThrottleCount metrics in Amazon CloudWatch for each API operation category at the account level. You can use CallCount to track the total number of calls customers made related to a category. You can use ThrottleCount to track the total number of throttled calls related to a category. You can use the CallCount and ThrottleCount metrics with the Sum statistic to count the total number of calls in a category. For more information, see CloudWatch usage metrics.

When monitoring service quotas, utilization is the percentage of a service quota in use. For example, if the quota value is 200 resources, and 150 resources are in use, the utilization is 75%. Usage is the number of resources or operations in use for a service quota.

Tracking usage through CloudWatch metrics

You can track and collect Amazon Cognito user pools utilization metrics with CloudWatch. The CloudWatch dashboard displays metrics about every AWS service that you use. With CloudWatch, you can create metric alarms to notify you or change a specific resource that you are monitoring. For more information about CloudWatch metrics, see Track your CloudWatch usage metrics.

Tracking utilization through Service Quotas metrics

Amazon Cognito user pools are integrated with Service Quotas, a console interface to display and manage your service quota usage. In the Service Quotas console, you can look up the value of a specific quota, view monitoring information, request a quota increase, or set up CloudWatch alarms. After your account has been active for a while, you can view a graph of your resource utilization.

The Utilization column in the Service Quotas console for Amazon Cognito user pools and Amazon Cognito identity pools displays your current quota. It also displays utilization for Amazon Cognito user pools. Adjustable Amazon Cognito user pools requests-per-second (RPS) quotas display their current usage. The Service Quotas console can also navigate you to CloudWatch metrics for a closer look at a selected quota metric. For more information on viewing quotas in the Service Quotas console, see Viewing Service Quotas.

Track monthly active users (MAUs)

The number of monthly active users (MAUs) in your user pool contributes important data to your planning for increases to request-rate quotas. You can compare your API request rates to the number of users you had active in a given time period. With that knowledge, you can calculate how an increase in active users of your applications will affect your quotas in your usage model. For example, imagine that your combined applications in US West (Oregon) resulted in 2 million active users in a month and your UserAuthentication category received occasional throttling errors at the default quota of 120 requests per second (RPS). In the previous month, before your successful advertising campaign, you had 1 million MAUs and your applications never exceeded 80 RPS. If you anticipate a similar spike as a result of a new TV spot, you might purchase an additional 40 RPS to accommodate the next million users with an adjusted quota of 160 RPS.

To review your MAUs

Access the AWS Billing console and review a recent bill. Under charges by service, you can filter on Cognito to view a breakdown of your MAUs for that billing period.

Requesting a quota increase

Amazon Cognito has a quota for the maximum number of operations per second that you can perform in your in the user pools and identity pools in each AWS Region. You can purchase an increase to adjustable Amazon Cognito user pools API request rate quotas. Check your current quota and purchase an increase from the Service Quotas console or with the Service Quotas API operations ListAWSDefaultServiceQuotas and RequestServiceQuotaIncrease.

  • To purchase a quota increase using the Service Quotas console, see Requesting a API quota increase in the Service Quotas User Guide.

  • AWS targets completion of quota increase requests within 10 days. However, several considerations might cause the request processing time to exceed 10 days. Some requests, for example, might require Amazon Cognito to provision additional hardware capacity, and seasonal increases in request volumes might introduce delays.

  • If the quota isn't available in Service Quotas, use the Service limit increase form.

Important

Only adjustable quotas can be increased. You must purchase increased quota capacity. For quota-increase pricing, see Amazon Cognito pricing.

Amazon Cognito user pools API operation categories and request rate quotas

Because Amazon Cognito has overlapping classes of API operations with differing authorization models, each operation belongs to a category. Each category has its own pooled quota for all member API operations, across all user pools in one AWS Region in your account. You can only request an increase to adjustable category quotas. For more information, see Requesting a quota increase. Quota adjustments apply to the user pools in your account in a single Region. Amazon Cognito restricts operations in some categories3 to 5 requests per second (RPS), per user pool. The Default quota (RPS) additionally applies to all user pools in an AWS account.

Note

The quota for each category is measured in Monthly Active Users (MAUs). AWS accounts with fewer than two million MAUs can operate within the default quota. If you have less than one million MAUs and Amazon Cognito is throttling requests, consider optimizing your app. For more information, see Optimize request rates for quota limits.

Category operation quotas are applied across all users in all user pools within one AWS Region. Amazon Cognito also maintains a quota for the number of requests that your app can generate against one user. You must limit per-user API requests as shown in the following table.

Amazon Cognito user pools per-user request rate quotas

Operation Operations per user per second
Read user profile

Examples: GetUser, GetDevice

10
Write user profile

Examples: UpdateUserAttributes, SetUserSettings

10

You must limit per-category API requests as shown in the following table.

Amazon Cognito user pools per-category request rate quotas

Category Description Default quota (RPS) Adjustable
UserAuthentication Operations that authenticate (sign in) a user.

These operations are subject to Amazon Cognito user pools API operations with special request rate handling .

120 Yes
UserCreation Operations that create or confirm an Amazon Cognito local user. This is a user that is created and verified directly by your Amazon Cognito user pools. 50 Yes
UserFederation

Operations that federate (authenticate) users with a third-party identity provider into your Amazon Cognito user pools.

Operations that submit an IdP response to a user pool federation endpoint. OIDC or social provider operations that result in an IdP token, and all SAML requests, contribute to this quota. 25 Yes
UserAccountRecovery Operations that recover a user's account, or change or update a user's password. 30 No
UserRead Operations that retrieve a user from your user pools. 120 Yes
UserUpdate Operations that you use to manage users and user attributes. 25 No
UserToken Operations for token management 120 Yes
UserResourceRead Operations that retrieve user resource information from Amazon Cognito, such as a remembered device or a group membership. 50 Yes
UserResourceUpdate Operations that update resource information for a user, such as a remembered device or a group membership. 25 No
UserList Operations that return a list of users. 30 No
UserPoolRead Operations that read your user pools. 15 No
UserPoolUpdate Operations that create, update, or delete your user pools. 15 No
UserPoolResourceRead Operations that retrieve information about resources, such as groups or resource servers, from a user pool.3 20 No
UserPoolResourceUpdate Operations that modify resources, such as groups or resource servers, in a user pool.3 15 No
UserPoolClientRead Operations that retrieve information about your user pool clients.3 15 No
UserPoolClientUpdate Operations that create, update, and delete your user pool clients.3 15 No
ClientAuthentication

client_credentials grant type requests to the token endpoint.

Operations that generate credentials to be used in authorizing machine-to-machine requests 150 No

1 AdminRespondToAuthChallenge and RespondToAuthChallenge are counted as UserAuthentication for NEW_PASSWORD_REQUIRED responses and UserAccountRecovery for all challenge responses other than NEW_PASSWORD_REQUIRED.

2 Each hosted UI operation during sign-in contributes one request to the quota. For example, a user who signs in and provides an MFA code contributes 2 requests. Token redemption in authorization-code grants is subject to an additional quota allocation at the same rate as your quota in the UserAuthentication category.

3 Any individual operation in this category has a constraint that prevents the operation from being called at a rate higher than 5 RPS for a single user pool.

Amazon Cognito identity pools (federated identities) API operation request rate quotas

Operation Description Default quota (RPS)1 Adjustable Quota increase eligibility
GetId Retrieve an identity ID from an identity pool. 25 Yes Contact your account team.
GetOpenIdToken Retrieve an OpenID token from an identity pool in the classic workflow. 200 Yes Contact your account team.
GetCredentialsForIdentity Retrieve AWS credentials from an identity pool in the enhanced workflow. 200 Yes Contact your account team.
GetOpenIdTokenForDeveloperIdentity Retrieve an OpenID token from an identity pool in the developer workflow. 50 Yes Contact your account team.
ListIdentities Retrieve a list of identity IDs in an identity pool. 5 Yes Contact your account team.
DeleteIdentities Delete one or more registered identities from an identity pool. 10 Yes Contact your account team.
TagResource Apply a tag to an identity pool. 5 Yes Contact your account team.
UntagResource Remove a tag from an identity pool. 5 Yes Contact your account team.
ListTagsForResource Display a list of the tags applied to an identity pool. 10 Yes Contact your account team.

1 The default quota is the minimum request rate quota for the identity pools in any AWS Region in your AWS account. Your RPS quota might be higher in some Regions.

Quotas on resource number and size

Resource quotas are the maximum number or size of resources, input fields, time duration, and other miscellaneous features in Amazon Cognito.

You can request an adjustment to some resource quotas in the Service Quotas console or from a Service limit increase form. To request a quota from the Service Quotas console, see Requesting a quota increase in the Service Quotas User Guide. If the quota isn't available in Service Quotas, use the Service limit increase form.

Note

Resource quotas at the AWS account level, like User pools per Region, apply to Amazon Cognito resources in each AWS Region. For example, you can have 1,000 user pools in US East (N. Virginia) and another 1,000 in Europe (Stockholm).

The following tables indicate default resource quotas, and whether they're adjustable.

Amazon Cognito user pools resource quotas

Resource Quota Adjustable Maximum quota
App clients per user pool 1,000 Yes 10,000
User pools per Region 1,000 Yes 10,000
Identity providers per user pool 300 Yes 1,000
Resource servers per user pool 25 Yes 300
Users per user pool 40,000,000 Yes Contact your account team.
Total combined changes in pre token generation Lambda trigger1 5,000 Yes Contact your account team.
Custom attributes per user pool 50 No N/A
Characters per attribute 2,048 bytes No N/A
Characters in custom attribute name 20 No N/A
Required minimum password characters in password policy 6–99 No N/A
Email messages sent daily per AWS account2 50 No N/A
Characters in email subject 140 No N/A
Characters in email message 20,000 No N/A
Characters in SMS verification message 140 No N/A
Characters in password 256 No N/A
Characters in identity provider name 32 No N/A
Identifiers per identity provider 50 No N/A
Identities linked to a user 5 No N/A
Callback URLs per app client 100 No N/A
Logout URLs per app client 100 No N/A
Scopes per resource server 100 No N/A
Scopes per app client 50 No N/A
Custom domains per account 4 No N/A
Groups to which each user can belong 100 No N/A
Groups per user pool 10,000 No N/A

1 This quota might be encountered in tokens from a Pre token generation Lambda trigger. The number of existing and added claims plus scopes in access and identity tokens must add up to a number smaller than or equal to this quota. Suppressed claims and scopes don't contribute to this quota.

2 This quota applies only if you are using the default email feature for an Amazon Cognito user pool. For a higher email delivery volume, configure your user pool to use your Amazon SES email configuration. For more information, see Email settings for Amazon Cognito user pools.

Amazon Cognito user pools session validity parameters

Token Quota
ID token 5 minutes – 1 day
Refresh token 1 hour – 3,650 days
Access token 5 minutes – 1 day
Hosted UI session cookie 1 hour
Authentication session token 3 minutes – 15 minutes

Amazon Cognito user pools code security resource quotas (non-adjustable)

Resource Quota
Sign-up confirmation code validity period 24 hours
User attribute verification code validity period 24 hours
Multi-factor authentication (MFA) code validity period 3–15 minutes
Forgot password code validity period 1 hour
Maximum number of ConfirmForgotPassword and ForgotPassword requests per user per hour1 5–20
Maximum number of ResendConfirmationCode requests per user per hour 5
Maximum number of ConfirmSignUp requests per user per hour 15
Maximum number of ChangePassword requests per user per hour 5
Maximum number of GetUserAttributeVerificationCode requests per user per hour 5
Maximum number of VerifyUserAttribute requests per user per hour 15

1 Amazon Cognito evaluates risk factors in the request to update passwords and assigns a quota that's tied to the evaluated risk level. For more information, see Forgot password behavior.

Amazon Cognito user pools user import job resource quotas

Resource Quota Adjustable Maximum quota
User import jobs per user pool 1,000 Yes Contact your account team.
Maximum characters per user import CSV row 16,000 No N/A
Maximum CSV file size 100 MB No N/A
Maximum number of users per CSV file 500,000 No N/A

Amazon Cognito identity pools (federated identities) resource quotas

Resource Quota Adjustable Maximum quota
Identity pools per account 1,000 No N/A
Amazon Cognito user pool providers per identity pool 50 Yes 1000
Character length of an identity pool name 128 bytes No N/A
Character length of a login provider name 2,048 bytes No N/A
Identities per identity pool Unlimited No N/A
Identity providers for which role mappings can be specified 10 No N/A
Results from a single list or lookup call 60 No N/A
Role-based access control (RBAC) rules 25 No N/A

Amazon Cognito Sync resource quotas

Resource Quota Adjustable Maximum quota
Datasets per identity 20 Yes Contact your account team.
Records per dataset 1,024 Yes Contact your account team.
Size of a single dataset 1 MB Yes Contact your account team.
Characters in dataset name 128 bytes No N/A
Waiting time for a bulk publish after a successful request 24 hours No N/A