View a markdown version of this page

Service quotas and throttling for Deadline Cloud - AWS Deadline Cloud
API request throttlingQuotas for related services

Service quotas and throttling for Deadline Cloud

AWS Deadline Cloud provides resources, such as farms, fleets, and queues, that you can use to process jobs. When you create your AWS account, we set default quotas on these resources for each AWS Region. Deadline Cloud also limits the rate of API requests. For more information, see API request throttling.

Service Quotas is a central location where you can view and manage your quotas for AWS services. You can also request a quota increase for many of the resources that you use.

To view the quotas for Deadline Cloud, open the Service Quotas console. In the navigation pane, choose AWS services and select Deadline Cloud.

To request a quota increase, see Requesting a quota increase in the Service Quotas User Guide. If the quota is not yet available in Service Quotas, use the service quota increase form.

Your AWS account has the following quotas related to Deadline Cloud.

The associated members quotas count the memberships that you assign to a farm, fleet, queue, or job. Each user grant and each group grant counts as one membership, so a group counts as one member no matter how many users it contains. Deadline Cloud doesn't limit the number of users in a group or the number of groups that a user belongs to; the quotas for AWS IAM Identity Center apply instead. To grant access to more users within the membership quota, assign groups instead of individual users. For more information, see How permissions work in Deadline Cloud.

Compute for service-managed fleets counts against the Deadline Cloud vCPU and GPU quotas in the following table, not against your Amazon Elastic Compute Cloud (Amazon EC2) service quotas. For more information, see Quotas for related services.

The following table includes quotas for persistent storage volumes used by service-managed fleets. For more information about persistent storage, see Persistent storage for service-managed fleets.

Name Default Adjustable Description
Associated members per farm Each supported Region: 75 No The maximum number of principals (users or groups) that can be associated to each farm in the current AWS Region. Each associated group counts as one member, regardless of how many users it contains.
Associated members per fleet Each supported Region: 75 No The maximum number of principals (users or groups) that can be associated to each fleet in the current AWS Region. Each associated group counts as one member, regardless of how many users it contains.
Associated members per job Each supported Region: 75 No The maximum number of principals (users or groups) that can be associated to each job in the current AWS Region. Each associated group counts as one member, regardless of how many users it contains.
Associated members per queue Each supported Region: 75 No The maximum number of principals (users or groups) that can be associated to each queue in the current AWS Region. Each associated group counts as one member, regardless of how many users it contains.
Budgets per farm Each supported Region: 20 Yes The maximum number of budgets per farm in the current AWS Region
Farms per region Each supported Region: 2 Yes The maximum number of farms that can be created in the current AWS Region.
Fleets per farm Each supported Region: 5 Yes The maximum number of fleets that can be created for each farm in the current AWS Region.
Jobs per farm Each supported Region: 100,000 Yes The maximum number of jobs per farm in the current AWS Region.
License endpoints per region Each supported Region: 5 Yes The maximum number of license endpoints in the current AWS Region.
License sessions per license endpoint Each supported Region: 500 Yes The maximum number of license sessions per license endpoint in the current AWS Region.
Limits per farm Each supported Region: 50 Yes The maximum number of limits that can be created for each farm in the current AWS Region.
Monitors per region Each supported Region: 1 No The maximum number of monitors in the current AWS Region.
OnDemand G instance GPUs per region Each supported Region: 0 Yes The maximum number of on-demand G instance GPUs that can be provisioned across all service-managed fleets in the current AWS Region.
OnDemand vCPUs per region Each supported Region: 50 Yes The maximum number of on-demand vCPUs that can be provisioned across all service-managed fleets in the current AWS Region.
Queue environments per queue Each supported Region: 10 No The maximum number of queue environments that can be created for each queue in the current AWS Region.
Queue fleet associations per farm Each supported Region: 100 Yes The maximum number of queue fleet associations per farm in the current AWS Region
Queue limit associations per queue Each supported Region: 10 Yes The maximum number of limits that can be associated with each queue in the current AWS Region.
Queues per farm Each supported Region: 20 Yes The maximum number of queues that can be created for each farm in the current AWS Region.
Resource configurations per fleet Each supported Region: 1 Yes The maximum number of VPC Lattice resource configurations that can be added to each fleet.
Spot G Instance GPUs per region Each supported Region: 0 Yes The maximum number of spot G instance GPUs that can be provisioned across all service-managed fleets in the current AWS Region.
Spot vCPUs per region Each supported Region: 50 Yes The maximum number of spot vCPUs that can be provisioned across all service-managed fleets in the current AWS Region.
Step consumers per step Each supported Region: 32 Yes The maximum number of steps that may declare a dependency on (consume) a single step within a job in the current AWS Region.
Step dependencies per step Each supported Region: 128 Yes The maximum number of steps that a single step within a job may declare a dependency on in the current AWS Region.
Steps per job Each supported Region: 200 Yes The maximum number of steps per job in the current AWS Region.
Storage for General Purpose SSD (gp3) volumes, in TiB Each supported Region: 1 Yes The maximum aggregated amount of EBS storage, measured in TiB, that can be used across all fleets in the current AWS Region.
Storage for persistent General Purpose SSD (gp3) volumes, in TiB Each supported Region: 4 Yes The maximum aggregated amount of storage, in TiB, that can be provisioned across persistent General Purpose SSD (gp3) volumes in the current AWS Region.
Storage profiles per farm Each supported Region: 50 No The maximum number of storage profiles that can be created for each farm in the current AWS Region.
Tasks per chunk Each supported Region: 150 No The maximum number of tasks that can be combined into a single chunk when submitting a job.
Tasks per job Each supported Region: 10,000 Yes The maximum number of tasks per job in the current AWS Region.
Tasks per step Each supported Region: 10,000 Yes The maximum number of tasks per step in the current AWS Region.
Wait-and-save vCPUs per region Each supported Region: 50 Yes The maximum number of wait-and-save vCPUs that can be provisioned across all service-managed fleets in the current AWS Region.
Workers per farm Each supported Region: 7,500 Yes The maximum number of workers per farm in the current AWS Region.

API request throttling

Deadline Cloud limits the rate of API requests for each AWS account in each AWS Region. The default request rates support large-scale workloads and usage. When your requests exceed the allowed rate, Deadline Cloud rejects the request with an HTTP 429 status code and a ThrottlingException error.

The AWS SDKs and the AWS CLI automatically retry throttled requests using exponential backoff. Occasional throttling resolves without requiring changes to your application. If you experience sustained throttling, contact AWS Support to request higher request rates.

Some Deadline Cloud features use other AWS services, and the quotas for those services also apply.

Amazon EC2 quotas for customer-managed fleets

Your Amazon Elastic Compute Cloud (Amazon EC2) quotas apply to customer-managed fleets, because those workers run on instances in your own account. Compute for service-managed fleets counts against the Deadline Cloud vCPU and GPU quotas in the preceding table instead.

To view or increase your Amazon EC2 quotas, open the Service Quotas console and choose Amazon Elastic Compute Cloud. For more information, see Amazon EC2 service quotas in the Amazon EC2 User Guide.

Amazon Bedrock quotas for the Deadline Cloud assistant

The Deadline Cloud assistant uses Amazon Bedrock on-demand inference in your AWS account, so your account's Amazon Bedrock service quotas apply. The two primary constraints are:

  • Requests per minute (RPM) – The number of model invocation requests allowed per minute.

  • Tokens per minute (TPM) – The total number of input and output tokens processed per minute.

Default quotas vary by Region. Some Regions have lower default limits, as low as 20 RPM, which might result in throttling during heavy assistant usage. The assistant uses cross-region inference profiles, which support a minimum of 200 RPM. Cross-region inference can help alleviate throttling in Regions with lower single-Region limits. If your Region uses cross-region inference, the service quotas in the destination Regions also apply.

If you experience throttling errors when using the assistant, you can request an Amazon Bedrock service quota increase:

To request a quota increase

  1. Open the Service Quotas console.

  2. In the navigation pane, choose AWS services, then choose Amazon Bedrock.

  3. Find the quota for the model used by the assistant (look for quotas related to InvokeModelWithResponseStream for the relevant model).

  4. Choose the quota name, then choose Request increase at account level.

  5. Enter your desired quota value and submit the request.

You can monitor your Amazon Bedrock quota usage through CloudWatch metrics. Set up CloudWatch alarms on Amazon Bedrock throttling metrics to identify when you are approaching your quota limits. For more information, see Monitoring Amazon Bedrock in the Amazon Bedrock User Guide.