Articles → AWS → Concurrency In AWS Lambda Function

Concurrency In AWS Lambda Function

In this article, we will discuss concurrency in AWS Lambda Function.

What Is Concurrency In Lambda Function?

Concurrency is the number of in-progress requests that your Lambda function can handle at the same time.

By default, the maximum concurrency limit is 1000 across all lambda functions across all regions.

Execution Environment

The execution environment executes the request. There are 2 phases of the execution environment: -

Init → In this phase, the execution environment got initialized
Invoke → In this phase, your lambda function is invoked

Picture showing the different phases of the execution environment

Click to Enlarge

Concurrency Request Lifecycle

For every concurrent request, the lambda function provisioned a new execution environment.

If there is a pre-initialized execution environment available and no request processing in progress then the request is processed by the execution environment else a new execution environment is created till the concurrency limit has been reached.

Click to Enlarge

How To Calculate Concurrency?

Concurrency = Average request per second * Average request duration in seconds

Consider the following example: -

Average request per second = 100

Average request duration in seconds = 0.5 seconds

Concurrency = 100* 0.5 = 50

It means that 50 execution environments are required to handle the request efficiently without throttle.

Types Of Concurrencies

There are 2 types of concurrencies: -

Reserved concurrency → A portion of your account’s concurrency is reserved for a function. All other functions use the available unreserved concurrency.
Provisioned concurrency → Several pre-initialized environments are allocated to the function

Difference Between The Reserved And Provisioned Concurrency

Reserved Concurrency	Provisioned Concurrency
In reserved concurrency, the allocated execution environments are not initialized. Every time, the lambda function initializes the execution environment and then uses it for function invocation. When the execution environment takes a long time than expected then it is called a Cold Start	In provisioned concurrency, the execution environments are pre-initialized
Lambda terminates the environment in case of inactivity	Lambda recycles the environment in the background
Function throttled when concurrency limit is reached	The function uses unreserved concurrency (other than the reserved instance) in case the concurrency limit is reached
No additional charges	Additional charges are incurred

Concurrency Quotas

The concurrency quota exists at two levels: -

Account level → As mentioned earlier, the concurrency quota at the account level is 1000
Function level → For all functions, you can reserve the maximum concurrency of 900. Pending 100; units are always reserved for the function that does not explicitly reserve concurrency

Posted By -	Karan Gupta

Posted On -	Thursday, January 26, 2023

Query/Feedback

Your Email Id		**

Subject		*

Query/Feedback	Characters remaining 250	**