Aurora Serverless platform version 4 changes scaling behavior

AWS announced Aurora Serverless platform version 4 on April 21, with changes to scaling behavior and performance. The release is relevant to teams running workloads where demand shifts between quiet periods and competing bursts of database activity.

Platform version joins the upgrade inventory

The announcement says new clusters, restored databases, and new clones launch on platform version 4. Existing clusters on versions 1 through 3 can upgrade through supported maintenance, stop-and-restart, or blue/green paths.

That introduces a practical comparison issue: two environments created at different times may not be running the same underlying platform version even when the team considers their database configuration equivalent.

Our analysis: record the platform version alongside the engine version when investigating a performance change. A reproducible test needs to identify both the application workload and the database environment it ran against.

Test transitions as well as steady load

Build a workload sequence that includes idle time, a sudden burst, overlapping jobs, and a return to normal demand. Observe latency and capacity behavior throughout the transitions instead of relying only on a steady-state throughput number.

Pay particular attention to the requests that arrive first after a quiet interval and to background work that competes with interactive traffic. Define what the application is allowed to retry and what users should see if a response takes longer than expected.

Treat the vendor’s performance claim as a hypothesis for evaluation, not an automatic budget or capacity reduction. Choose an upgrade route that fits the team’s interruption tolerance, preserve recovery evidence, and compare the same workload before and after the change. Expand only when the measured transition behavior meets the service’s objectives.