Global client

[valeriomazzeo]

Both cache and rateLimit options are worthless if the client is recreated for each lambda execution.

Due to the stateless nature of lambda functions, this is not able to ensure proper caching, but at least on hot starts the client will be reused for each lambda container which is better than nothing.

Am I missing anything? Is the client cache somehow working even when recreated each time?

[agustin-vieta]

@valeriomazzeo
Some form of caching can be achieved via AWS Lambda execution context:

After a Lambda function is executed, AWS Lambda maintains the execution context for some time in anticipation of another Lambda function invocation.

To use that, your variables must be declared outside of the function's handler:

Objects declared outside of the function's handler method remain initialized, providing additional optimization when the function is invoked again. For example, if your Lambda function establishes a database connection, instead of reestablishing the connection, the original connection is used in subsequent invocations.

However, the problem with using the execution context is that AWS does not provide any guarantee on how long you can use it:

When you write your Lambda function code, do not assume that AWS Lambda automatically reuses the execution context for subsequent function invocations.

Thus, the caching solution based on lambda execution context could work, but it would be limited in many ways. For proper caching solutions you would have to look outside of lambda, for example store your serialized objects in ElastiCache, DynamoDB or more.