AWS Aurora Support
Aurora Serverless v2 capacity (ACU) tuning.
Serverless v2 scales between a minimum and maximum number of Aurora capacity units (ACUs). Set the minimum too low and spikes arrive before capacity does. Set it too high and you pay for idle memory. We size both ends from your actual load pattern.
When You Need This
The situations where this work pays for itself.
Signs it is time
- Your Serverless v2 bill is higher than expected
- Queries slow down right after traffic spikes
- The cluster rarely scales down or rarely scales up
- Connection limits or memory errors appear at low capacity
- You are deciding between Serverless v2 and provisioned instances
What you get
- Capacity analysis with recommended min and max ACU
- Cost comparison against provisioned instances
- Alarms on capacity at maximum and on memory pressure
- Infrastructure as code for the new settings
How we do it
A clear, safe process you can follow along with.
1. Collect the capacity history
We plot ServerlessDatabaseCapacity, ACUUtilization, CPUUtilization, FreeableMemory and DatabaseConnections over several weeks to see baseline, peaks and how fast demand rises.
2. Set the minimum from the baseline
Minimum capacity decides the floor of memory and buffer cache. A very low minimum means a cold cache after quiet periods and slower scale-up from a small base. We set it so common peaks do not start from scratch.
3. Set the maximum as a cost and safety cap
The maximum bounds both performance and spend. We pick it from the observed peak plus headroom and alarm when the cluster sits at the cap, which signals real saturation.
4. Check connection and memory limits
Some engine limits, such as max_connections, depend on the configured capacity. We verify the workload fits at the minimum and does not hit memory pressure when scaled down.
5. Review scaling behavior and the writer/reader mix
Each instance in the cluster scales separately, and promotion tiers affect readers. We review whether readers should scale with the writer and use a provisioned writer if load is steady.
6. Compare with provisioned cost
For steady high load, provisioned instances with reservations are often cheaper. We model both using your measured ACU hours and current pricing.
7. Apply and monitor
We change settings in infrastructure as code, run a representative load, and review capacity and latency over the following weeks.
Mistakes we help you avoid
Common ways this goes wrong, and what we do instead.
Minimum set to the lowest value for every environment
Production usually needs a higher floor so cache and connections hold up.
No maximum alarm
Sitting at max capacity looks like an app problem, but the cap is the cause.
Using Serverless v2 for a flat workload
A steady load may be cheaper on provisioned instances.
Frequently Asked Questions
What is an ACU in Aurora Serverless v2?
An Aurora capacity unit combines compute and memory. Serverless v2 scales the number of ACUs up and down within your minimum and maximum settings.
Can Aurora Serverless v2 scale to zero?
Newer versions support automatic pause with a minimum of zero ACUs for some engine versions. Check the current documentation for your version and the resume delay.
Why is my Serverless v2 database slow after idle time?
A low minimum shrinks the buffer cache, so the first queries after quiet periods read from storage. Raising the minimum keeps more cache warm.
Want Help With Aurora Serverless v2 capacity (ACU) tuning?
Schedule an appointment. Tell us about your environment and we will outline the safest, highest-value approach.
Schedule Appointment › No commitment required