Explanation
Why the waste happens and who it affects.
Enterprise Plus adds a 99.99% availability SLA that includes maintenance, sub-second maintenance downtime, the optional data cache, larger machine types and up to 35 days of point-in-time recovery. Enterprise offers a 99.95% SLA (excluding maintenance) and up to 7 days of point-in-time recovery, and Google positions it as balanced performance at the lowest cost. vCPU and memory are priced higher on Enterprise Plus, and the premium applies to every hour the instance runs.
Many instances land on Enterprise Plus without a deliberate choice. It is the default edition for new PostgreSQL 16 and later instances and for MySQL 8.4 and 9.7, so dev, test, internal tools and modest production databases created with defaults pay the premium. Teams also upgrade for a specific performance problem and never revisit the decision. For workloads that do not need the higher SLA, near-zero-downtime maintenance, data cache or long point-in-time recovery, the extra spend buys nothing.
Billing model
The pricing dimensions that drive this cost.
Edition sets the per-hour rate for vCPU and memory; rates vary by region and machine series.
- Enterprise Plus vCPU and memory
- Billed per vCPU-hour and GiB-hour at higher rates than Enterprise; in us-central1 the published N2 rates are about 30% above Enterprise general purpose rates
- Data cache
- Optional Enterprise Plus feature billed per GiB-hour of cache storage in addition to vCPU and memory
- HA multiplier
- HA instances pay double vCPU, memory and data cache rates in either edition, so the edition premium doubles too
- Committed use discounts
- Cloud SQL CUDs apply to vCPU and memory in both editions but do not remove the price gap
How to detect
4 checks to find it in your estate.
- List instances with their edition, database version, machine type and environment (gcloud sql instances list with settings.edition) and flag Enterprise Plus instances in non-production projects
- Flag Enterprise Plus instances that do not use data cache, run at low CPU and memory utilization, and have no documented need for the 99.99% SLA or sub-second maintenance downtime
- Check point-in-time recovery retention (transactionLogRetentionDays); instances that keep 7 days or less do not use the longer Enterprise Plus window
- Review provisioning templates for PostgreSQL 16+ and MySQL 8.4+ instances that omit the edition and therefore get Enterprise Plus by default
How to fix
4 ways to remove the waste.
- Switch suitable instances to Enterprise edition in place with gcloud sql instances patch INSTANCE --edition=enterprise --tier=MACHINE_TYPE, or Switch to Enterprise in the console; the endpoint does not change, but downtime is longer than for the upgrade direction
- Choose a general purpose dedicated-core or N4 machine type for the target; shared-core machine types are not available when downgrading, and moving from N4, C4 or C4A back to a general purpose or N2 series migrates Hyperdisk Balanced storage to SSD, which can incur significant downtime
- Confirm the workload fits Enterprise limits (up to 96 vCPUs and 624 GiB memory) and that 7 days of point-in-time recovery and the Enterprise SLA are acceptable; the data cache is dropped
- Set the edition explicitly in Terraform and scripts so new PostgreSQL 16+ and MySQL 8.4+ instances do not default to Enterprise Plus
Documentation
Vendor references for pricing and configuration.