How does Lambda@Edge scaling work?

Lambda@Edge capacity scales dynamically in response to increased traffic. It depends on the available concurrency of AWS Lambda in your account and AWS regions. While the concept of a region doesn’t apply to Lambda@Edge functions, regional AWS Lambda quotas correspond with Lambda@Edge scaling in Regional Edge Caches. The following quotas apply:

• for requests per second, per region, per function
• for concurrent executions per region, across all functions in each AWS account

These quotas can be increased by requesting a service quota increase. Additionally, the following quotas apply:

• for additional execution environments provisioned every 10 seconds, per region, per function
• for requests per second, per execution environment, even if the function execution takes less than 100ms

While the two quotas regarding execution environments can’t be changed for on-demand capacity, you are able to use AWS Lambda reserved concurrency and provisioned concurrency options for Lambda@Edge functions.

With reserved concurrency, you can reserve a portion of your account's concurrency for a specific function, to ensure that critical functions always get the concurrency they need.

With provisioned concurrency, execution environments are ready to respond immediately to incoming function requests. This is useful not only for the sake of providing concurrency to handle web traffic, but also for reducing cold start latencies of functions.

For more information, see Function scaling in the AWS Lambda Developer Guide.