Service quotas

Service quotas, also referred to as limits, are the maximum number of service resources or operations for your AWS account.

Request a service quota increase based on anticipated usage

DeepRacer on AWS uses a combination of Amazon SageMaker AI training jobs and AWS Lambda functions for servicing compute-related tasks, including but not limited to creating, training, and evaluating models. AWS Lambda functions are also used for servicing general requests, such as creating a new account, making changes to account settings, and running/participating in races.

Considering your anticipated usage in advance and right-sizing your service quotas can improve user experience by reducing the amount of time it takes for training and evaluation jobs to be run. If your deployment is expected to serve consistently high demand (i.e. large number of users) or is subject to burst traffic (i.e. used for events, classes, workshops etc.), training and evaluation jobs will be constrained by the service quota that is in place, and those jobs will remain in the queue until capacity is available to run them.

Amazon SageMaker AI training jobs

Amazon SageMaker AI training jobs are responsible for running training and evaluation jobs, and also support community races. At the time of writing, the default applied account-level quota value is 8. This means that DeepRacer on AWS can dispatch up to 8 training jobs, evaluation jobs, or race submissions (i.e. evaluating one model in a given race) at a time. If demand exceeds this limit at any point, jobs will remain queued until an actively-running job is completed and capacity becomes available. The queue for jobs in DeepRacer on AWS is managed in FIFO (first-in-first-out) order, and the amount of time that a job spends in the queue depends on the number of jobs that can be processed currently (i.e. the service quota), and the number of jobs that have entered the queue ahead of it.

As a result, it is recommended to consider in advance the number of jobs that may need to be run concurrently at any given point in time in order to deliver preferable throughput. If you decide that you would like to request a service quota increase, you may do so by:

AWS Lambda functions

AWS Lambda functions are the primary compute resource used for servicing requests in DeepRacer on AWS, including dispatching and monitoring training and evaluation jobs. Similar to that of the Amazon SageMaker AI training jobs noted in the previous section, AWS Lambda functions have an applied account-level quota of 1,000 concurrent executions. This represents the maximum number of events that functions can process simultaneously in the current region. In the event that this limit is reached, requests will be queued and serviced once capacity becomes available.

If you anticipate needing to be able to service more than 1,000 requests concurrently, requesting a service quota increase for the number of requests that you would like to be able to handle at any given point is recommended. This will allow requests to be serviced as soon as they are received, or with minimal wait time in the queue. If you decide that you would like to request a service quota increase, you may do so by:

Amazon Virtual Private Cloud (VPC)

DeepRacer on AWS configures one Amazon Virtual Private Cloud (VPC) per deployment to provide network isolation for functions that handle imported models and other user-supplied artifacts. At the time of writing, the default account-level quota is 5 VPCs per region. If you plan to host more than 5 deployments of DeepRacer on AWS in a single region, requesting a service quota increase for at least the number of deployments you expect to host is recommended.

Quotas for AWS services in this solution

Make sure you have sufficient quota for each of the services implemented in this solution. For more information, see AWS service quotas.

Use the following links to go to the page for that service. To view the service quotas for all AWS services in the documentation without switching pages, view the information in the Service endpoints and quotas page in the PDF instead.