StatusNeo

AWS LAMBDA COLD START

Introduction (What is Cold Start)

• When a serverless function is running, it stays in an active state only as long as it is in the running state.
• Containers also remain in an active state only during this duration of time.
• After some time these containers become inactive if no activity is there.
• Functions will go in an inactive state if no execution is there.
• Inactivation of our functions or containers is done by the cloud provider that we are using say for example – AWS or GCP.
• When we invoke an inactive serverless function that start is called a Cold Start.
• Delay occurs because the function or the container is in an inactive state, and it takes some time for the cloud provider to restore their state.
• Spinning up a new container after it has been in an inactive state causes the delay.
• This is how cold start affects the execution time of a serverless function.
• In short, we can say that cold start results in a longer execution time of the serverless functions resulting in the overall delay of the application.

• In the above diagram, the first two steps are the cold start duration.
• Firstly, the download of our code happens then a new environment is set up.
• Although we are not charged for this time duration it causes a delay in the execution of our function.
• Once the execution is completed, the execution environment is frozen for a certain period of time.
• If another request arrives for the same function, the service may reuse the environment that is frozen to avoid latency and for better performance.
• This use of the previous environment is called a warm start as it doesn’t download your code and set up the environment again.
• According to research cold start invokes only under 1% of the invocations.
• The duration of cold start varies from 100ms to up to 1 sec.

Reasons for the Cold Start

• Whenever the function is updated or changed there will always be a cold start as it reloads the code and creates a new environment.
• When the function is not invoked frequently.
• If the programming language in which lambda is used has a higher latency or more dependencies.

Reducing Cold Start

• Provisioned Concurrency can be used if we need a predictable function start time.
• It means to increase the amount of concurrency during high demand and decrease it at the time of low demand.
• It can be done through AWS CLI and no need to make any changes to your code.
• Deciding the right memory size can also be helpful in reducing the cold start.
• Monitoring the performance of the lambda and setting up the configuration according to that only will also help.
• Caching can be used if the invocation is not very frequent or the data is not getting updated very frequently.
• Low Latency languages can be used to avoid delay while the execution.

Conclusion

• Cold Start is something that will take place.
• We need to understand our needs and define our configuration according to that.
• Provisioned Concurrency is one of the best ways to optimize the cold start.
• Better memory management is also very helpful.

References

• https://aws.amazon.com/blogs/compute/operating-lambda-performance-optimization-part-1/

Still Curious? Visit my website to know more!

For more interesting Blogs Visit- Utkarsh Shukla Author