[Feature Request] Better integration with RDS

Right now the only official way to integrate with RDS is by using the Data API. The Data API is only available with Aurora Serverless V1, which has been deprecated. Serverless V2 and most other deployment types require a SQL connection either directly or through RDS Proxy.

We began using AppSync 2 years ago, along with DynamoDB. NoSQL fit our data requirements at the time, but those requirements have evolved and we will have to migrate to a relational database. At this point, AppSync's lack of compatibility with RDS is holding us back and we would love to see it added.

One solution we are thinking of, and which is mentioned in a previous issue #270, is using Lambdas to forward the SQL queries. However, Lambda cold start times result in latency that is significant enough to be noticeable by end users (and frustrate my boss). You have to pay for both AppSync and Lambda to sit there and wait for the database to respond, and Lambda concurrency limits create a scaling bottleneck. We will probably have to implement this soon but are really hoping other solutions will be built before we hit this bottleneck.

Another possible solution that may be easier for you guys to implement would be adding VPC support to the HTTP data source. While not as good as a direct RDS connection, we'd be able to host an intermediate layer on ECS or Fargate which are more suitable and scalable for this type of usage. As it is now, if we went this route we'd have to expose our intermediate layer to the internet.

No repository files, tests, or entry points are identified. Start by reading issue #270 and comparing the proposed Lambda, VPC HTTP data source, ECS/Fargate, and direct RDS approaches described here. Done would require a decided scope and an implemented, scalable AppSync-to-RDS integration path.

Written by the indexing model from the issue text.