[reilly3000]

Aurora's autoscaler is fully in your control. Aurora Serverless is usage-based, but regular Aurora is just an AWS-optimized Postgres distro. You should not receive any surprise billing based on scaling, unless you specifically configure it to do so. Based on those usage patterns you explained you could do a scheduled scaling event that would scale up your Aurora cluster to have more capacity during business hours. If its a read-heavy workload then read replicas are a cheap and easy way to scale.

[orf]

‘AWS optimized Postgres distro’ is perhaps underselling it a bit, while being broadly correct. The magic is in the storage engine which powers instantaneous failover, point in time restores and the newly released ‘serverless 2’ product (which would be ideal for the parent has)

[reilly3000]

You're spot on that Aurora is super-powerful for performance and scaling, giving it characteristics and controls that don't really exist elsewhere. I heard a great talk with their engineering lead about how they optimized Postgres internals to play nicely with S3.

I still think a small Aurora Postgres with some workday scaling events on a schedule would have more affordable and predictable if the bulk of the workload is 9-6 weekdays. Serverless Aurora is (finally) great, but expensive and really requires some extra work to optimize usages to minimize cost.