The Best practices blog for Aurora Serverless v1 says:
you should be mindful of a few things, such as connection management and cold starts. ... Serverless applications can open a large number of database connections or frequently open and close connections. This can have a negative impact on the database and lead to slower performance. ... If there is a sudden spike in requests, you can overwhelm the database. Aurora Serverless v1 might not be able to find a scaling point and scale quickly enough
We'd like to find a way to speed scaling up on high load, though the
hey load tester is not realistic. We'd like it to scale so we
don't run out of database connections. This doc discusses How Aurora
Serverless v1 works
So why not use Aurora Serverless v2, instead of v1? We're using v1 because it scales to 0 ACU, so costs nothing but the disk our data needs. Aurora Serverless v2 scales down to 0.5 ACU, which means it costs money for CPU even if nothing is using it -- exactly what we are trying to avoid here.
But v2 scales up much faster than v1. This Serverless Guru tip provides a good comparison; briefly:
- Aurora v1 scales up in seconds, v2 in milliseconds
- v1 can take 15 minutes to scale down, v2 in under a minute
- v2 scales much more finely, in 0.5 ACU increments.
The Performance and scaling doc goes into detail on this.