Explanation
Why the waste happens and who it affects.
Many PostgreSQL workloads do not need high sustained CPU and can run on fewer vCores. Memory-optimized (E-series) instances reduce cost only when fewer vCores with more memory per vCore replace a larger general-purpose server, and burstable (B-series) instances suit development, test, or low-concurrency workloads.
When actual usage consistently falls below the provisioned capacity - particularly CPU - the deployment may be overprovisioned, resulting in unnecessary compute charges. Choosing the wrong VM family or size leads to silent overspend, especially in long-lived database environments.
Billing model
The pricing dimensions that drive this cost.
- Charged based on:
- Provisioned compute (vCores and instance family)
- Provisioned storage (per GB/month)
- Pricing varies significantly by VM series (e.g., D-series vs. B-series vs. E-series)
- Charges accrue continuously, regardless of actual CPU or memory usage
How to detect
4 checks to find it in your estate.
- Analyze average CPU and memory utilization of PostgreSQL Flexible Server instances
- Determine whether current performance requirements justify the provisioned instance size
- Identify servers running on D-series or other general-purpose SKUs that could be supported by B-series or E-series
- Validate compatibility and performance before considering a shift in instance family
How to fix
4 ways to remove the waste.
- Resize the PostgreSQL Flexible Server to a smaller or more suitable VM family based on actual workload behavior
- For low-CPU workloads, first reduce vCores within the same family. Consider E-series (memory-optimized) only if you can cut vCores while preserving memory, since E-series costs more per vCore than D-series, and keep B-series (burstable) for dev/test or low-concurrency workloads, as Microsoft does not recommend Burstable for production
- Review usage patterns quarterly to ensure the selected SKU remains aligned with performance needs
- Avoid unnecessary performance buffer unless justified by workload spikes or growth
Documentation
Vendor references for pricing and configuration.