Aurora Serverless: Data API vs Connection pools

Amazon Aurora Serverless is the goto relational database choice for Serverless application on AWS.

Developers usually confront two problems when using Aurora Serverless.

• Connection pool management is really hard when using Lambda’s. A flood of requests from quickly scaling Lambda’s can bring the database to it’s knees.
• The Data API is too slow. With latencies around 150 to 250 ms for even a simple request.

So which method do you use to connect to your database?

The answer is both.

Use connection pools and precious direct database connections to handle user facing API’s and time critical operations.

Use the Data API for background/scheduled processes, running database migrations or access data that can be lazy loaded in an app.

In the example architecture above, which represents a blogging platform, we use connection pools and a direct connection to the database to create and fetch a blog post. This requires lambda’s to be inside a VPC, along with the RDS database.

Use the Data API to lazy load all the comments for the blog post. The Data API can be accessed from outside the VPC.