AWS RDS vs Google Cloud SQL: Engines, HA, Pricing (2026)

Quick verdict

Short answer

Choose Amazon RDS if your applications run on AWS, or if you need MariaDB, Oracle or Db2, which Cloud SQL does not offer. Choose Cloud SQL if you are on Google Cloud and want MySQL, PostgreSQL or SQL Server with a simple edition choice: Enterprise for general workloads, or Enterprise Plus when you need a 99.99% SLA that includes maintenance, sub-second planned maintenance downtime and up to 35 days of point-in-time recovery. For the same engine, the features are close enough that the cloud your applications live in should usually decide.

How we know: This comparison is research-based: engines, versions, editions, high availability, backups, read replicas, maintenance and pricing models were checked against the AWS RDS product and pricing pages, the Amazon RDS User Guide and Google's Cloud SQL documentation in October 2026. Google's Cloud SQL pricing page could not be read in full, so no Cloud SQL list prices are quoted. We have not run performance or cost tests.

Amazon RDS (Amazon Relational Database Service) is AWS's managed relational database service. The RDS User Guide lists six engines: IBM Db2, MariaDB, Microsoft SQL Server, MySQL, Oracle Database and PostgreSQL. You create a DB instance of a chosen instance class and storage type, then decide on Multi-AZ, read replicas and backup retention. AWS handles backups, patching, failure detection and recovery. The AWS RDS product page also groups the Aurora family under RDS, but Aurora is a separate architecture (see Amazon RDS vs Amazon Aurora). RDS Custom, for Oracle and SQL Server only, adds operating system access.

Cloud SQL is Google Cloud's managed service for MySQL, PostgreSQL and SQL Server. Each instance belongs to one of two editions, Cloud SQL Enterprise or Cloud SQL Enterprise Plus, and both editions support all three engines. The edition sets the SLA, the available machine types, planned maintenance downtime, the point-in-time recovery window and features such as a data cache and read pools. For PostgreSQL workloads that need more than Cloud SQL offers, Google sells AlloyDB for PostgreSQL separately; see Google Cloud SQL vs AlloyDB.

Side by side

Aspect	Amazon RDS	Cloud SQL
Engines	Db2, MariaDB, SQL Server, MySQL, Oracle, PostgreSQL	MySQL, PostgreSQL, SQL Server
Tiering	Instance classes and storage types; HA and replicas are per-instance options	Two editions: Enterprise and Enterprise Plus
MySQL and PostgreSQL versions	Several major versions of each, listed per engine in the RDS User Guide; Extended Support charges after community end of life	MySQL 5.6, 5.7, 8.0, 8.4 (default), 9.7; PostgreSQL 9.6 to 18 (18 default), with 12 and earlier in paid extended support
SQL Server versions	2016, 2017, 2019, 2022, 2025	2017, 2019, 2022, 2025 (no Web edition on 2025)
High availability	Multi-AZ DB instance (standby not readable) or Multi-AZ DB cluster with two readable standbys (MySQL, PostgreSQL)	Regional instance: primary and standby in two zones with synchronous disk replication; standby not readable
SLA by tier	No edition-based tiers; availability depends on the Multi-AZ option you choose	Enterprise 99.95% (excludes maintenance); Enterprise Plus 99.99% (includes maintenance)
Automated backups	Retention 0 to 35 days; point-in-time restore within that window	Scheduled automated backups; retention up to 10 years with enhanced backups; PITR up to 7 days (Enterprise) or 35 days (Enterprise Plus)
Read scaling	Asynchronous read replicas, including cross-Region; up to 15 per source for SQL Server (Enterprise edition)	Read replicas, cross-region replicas; read pools and cascading replicas in Enterprise Plus
Planned maintenance	Weekly 30-minute window; Multi-AZ reduces OS patching to a failover	One-hour window you choose; under 60 seconds downtime (Enterprise) or under 1 second (Enterprise Plus)
Maximum storage	Up to 64 TiB, depending on engine and storage type (SQL Server: 16 TiB General Purpose SSD, 64 TiB Provisioned IOPS)	Up to 64 TB on dedicated-core instances; 3 TB on shared-core
Main trade-off	More engines, more options to configure and price separately	Fewer engines; the best maintenance and PITR terms require Enterprise Plus

Key differences

Engines and editions

Both services run the three engines most teams compare: MySQL, PostgreSQL and SQL Server. RDS also runs MariaDB, Oracle Database and IBM Db2 (Standard, Advanced and Community editions), which matters if you have commercial databases to move. Cloud SQL has no Oracle, MariaDB or Db2 option.

Cloud SQL publishes its version list clearly: MySQL 5.6, 5.7, 8.0, 8.4 (the default) and 9.7; PostgreSQL 9.6 through 18 (18 is the default), with versions 12 and earlier in extended support that is charged; and SQL Server 2017, 2019, 2022 and 2025. Google says Cloud SQL supports new PostgreSQL major versions within 90 days of community release, and that new instances can no longer be created on SQL Server 2017 from 12 April 2027. RDS documents versions per engine; for SQL Server it lists 2016 to 2025, and AWS charges Extended Support for engines past community end of life.

The tiering model differs. On RDS you pick an instance class and storage type and then add features. On Cloud SQL you first pick an edition. Enterprise Plus adds a 99.99% SLA that includes maintenance, larger machines (up to 128 vCPUs and 864 GB of memory on N2), a data cache, read pools, planned maintenance downtime of under one second and up to 35 days of point-in-time recovery, against 99.95% excluding maintenance, under 60 seconds and up to 7 days for Enterprise.

High availability

RDS offers two Multi-AZ shapes. A Multi-AZ DB instance keeps a synchronous standby in another Availability Zone that does not serve reads; for SQL Server it uses Always On availability groups or database mirroring. A Multi-AZ DB cluster (MySQL and PostgreSQL) has a writer and two readable standbys across three zones, using semisynchronous replication, and AWS says it typically has lower write latency than a Multi-AZ DB instance.

Cloud SQL HA uses a regional instance: a primary and a standby in two zones, with every write replicated synchronously to disks in both zones before the transaction is reported as committed. Google says to expect about sixty seconds of unavailability during a failover, and the standby cannot serve reads. Google also states that an HA instance costs twice as much as a standalone instance, covering CPU, memory and storage.

Backups and point-in-time recovery

RDS takes storage-volume snapshots of the whole instance during a backup window, keeps automated backups for a retention period of 0 to 35 days (the console default is 7, the API default 1), and supports point-in-time restore within that period. Manual snapshots are separate, kept until you delete them, and limited to 100 per Region. Backups are stored in Amazon S3, and you can retain automated backups when you delete an instance.

Cloud SQL takes incremental automated backups on a schedule and on-demand backups. It now offers two backup options: standard backups stored in the same project, and enhanced backups managed centrally through the Backup and DR service with enforced retention up to 10 years. Point-in-time recovery depends on log retention, which is up to 7 days in Enterprise and up to 35 days in Enterprise Plus. Cloud SQL can also take a final backup when an instance is deleted.

Read replicas and scaling

RDS read replicas use each engine's native asynchronous replication and are created manually; RDS does not autoscale them. MySQL and MariaDB replicas can be made writable, and MySQL, MariaDB and some PostgreSQL versions support replicas of replicas. For SQL Server, read replicas require Enterprise edition, an instance class with at least four vCPUs, and are limited to 15 per source. Oracle and Db2 replicas can run in mounted or standby mode for cross-Region disaster recovery.

Cloud SQL supports read replicas and cross-region replicas for its engines; for SQL Server they require Enterprise edition, up to 8 per primary, and cascading replicas need Enterprise Plus. Enterprise Plus also adds read pools. Google charges a replica at the same rate as a standard instance, with no charge for the replication itself within a region.

Maintenance

RDS gives every instance a weekly 30-minute maintenance window. Required OS and database patches are applied in it; AWS says such patching is infrequent, typically every few months, and that mandatory upgrades can be applied after their apply date even if you move the window. Multi-AZ turns OS patching into a short failover, and Blue/Green Deployments can reduce upgrade downtime for MySQL, MariaDB and PostgreSQL.

Cloud SQL maintenance also happens about every few months. You choose a one-hour window and a timing preference within a rollout, can set deny maintenance periods of up to 90 days, receive email notice at least a week ahead, and can apply maintenance yourself early. Google states connectivity loss of under 60 seconds on average for Enterprise and under 1 second for Enterprise Plus.

Pricing and licensing

Amazon RDS. According to the AWS RDS pricing page in October 2026, you pay for DB instance hours by instance class, provisioned storage per GB-month, Provisioned IOPS where used, backup storage, and data transfer between Availability Zones and Regions. Partial instance hours are billed per second with a 10-minute minimum. Reserved instances for one or three years advertise savings of up to 69%, and Database Savings Plans are also listed. Multi-AZ, read replicas (billed as normal instances) and licence-included engines such as SQL Server and Oracle raise the cost, and engines past community end of life incur Extended Support charges. New AWS customers get up to USD 200 in Free Tier credits on a 6-month free plan; RDS Free Tier covers MariaDB, MySQL, PostgreSQL and SQL Server Express on micro instances, Single-AZ only.

Cloud SQL. Cloud SQL bills instance CPU and memory, storage, backups, IP addresses, network egress and, for SQL Server, licensing. Google's committed use discount documentation lists 25% off for a 1-year commitment and 52% for 3 years, applied to CPU and memory in the committed region across Enterprise and Enterprise Plus, but not to shared-core machine types, storage, backups, IP addresses, network egress or licensing. An HA instance costs twice a standalone instance, and read replicas are charged as standard instances. PostgreSQL 12 and earlier incur extended support charges. We could not read Google's pricing page reliably, so no list prices are quoted; use the Google Cloud pricing calculator.

Both bills depend on engine, edition or instance class, HA, region, storage and commitment, so compare like-for-like configurations in each vendor's calculator.

Pricing checked on the vendors' official pages on 7 October 2026. Prices change; confirm before buying.

Where each one leads

Amazon RDS strengths

Six engines, including MariaDB, Oracle and Db2, which Cloud SQL does not offer
Multi-AZ DB clusters with two readable standbys for MySQL and PostgreSQL
Backup retention up to 35 days on every instance, without an edition upgrade
RDS Custom for Oracle and SQL Server when you need OS-level access
Blue/Green Deployments to reduce upgrade downtime for MySQL, MariaDB and PostgreSQL

Cloud SQL strengths

Simple two-edition model with clearly published SLA, maintenance and PITR terms
Enterprise Plus: 99.99% SLA including maintenance and sub-second planned maintenance downtime
Recent engine versions documented clearly, with PostgreSQL 18 as the default
Deny maintenance periods of up to 90 days and self-service maintenance
Enhanced backups with centrally enforced retention of up to 10 years

Limitations

Amazon RDS limitations

Many options (instance class, storage type, Multi-AZ shape, replicas) to choose and price separately
The Multi-AZ DB instance standby cannot serve reads
No read replica autoscaling; SQL Server replicas need Enterprise edition
No sysadmin or host access on standard RDS

Cloud SQL limitations

No Oracle, MariaDB or Db2
HA standby cannot serve reads, and HA doubles the instance cost
Point-in-time recovery is limited to 7 days on Enterprise; 35 days needs Enterprise Plus
SQL Server instances restrict sysadmin and lack CLR, FILESTREAM, Database Mail and SSAS

When to choose each

Choose Amazon RDS if

Your applications run on AWS
You need managed Oracle, MariaDB or Db2
You want readable HA standbys for MySQL or PostgreSQL (Multi-AZ DB cluster)
You need operating system access for Oracle or SQL Server through RDS Custom

Choose Cloud SQL if

Your applications run on Google Cloud
You want a published 99.99% SLA that includes maintenance (Enterprise Plus)
You need tight control of maintenance timing, including deny periods
You want the newest PostgreSQL or MySQL major versions as soon as Google supports them

When neither is right

You need PostgreSQL with storage that scales automatically or serverless compute: on AWS compare Amazon RDS vs Amazon Aurora; on Google Cloud see Google Cloud SQL vs AlloyDB.
You want serverless Postgres with branching for development: see Neon vs AWS RDS and Supabase vs Neon.
You run SQL Server and are open to Azure: see AWS RDS vs Azure SQL Database and Azure SQL vs Google Cloud SQL.
You need full control, sysadmin or OS access for every engine: run the database yourself on virtual machines.

Final recommendation

Bottom line

For MySQL, PostgreSQL and SQL Server, Amazon RDS and Cloud SQL cover the same core jobs: managed backups, point-in-time recovery, zone-level HA, read replicas and scheduled maintenance. Pick the one in the cloud where your applications run. RDS is the stronger choice when you need Oracle, MariaDB or Db2, readable HA standbys, or 35-day backups without an edition upgrade. Cloud SQL is simpler to reason about, and Enterprise Plus offers maintenance and SLA terms that RDS only approaches through Multi-AZ configuration.

Frequently asked questions

Does Google Cloud SQL support Oracle or MariaDB?

No. Cloud SQL supports MySQL, PostgreSQL and SQL Server. Amazon RDS supports those three plus MariaDB, Oracle Database and IBM Db2.

What is the difference between Cloud SQL Enterprise and Enterprise Plus?

Per Google's documentation, Enterprise Plus has a 99.99% SLA that includes maintenance (Enterprise: 99.95% excluding maintenance), larger machine types, a data cache, read pools, planned maintenance downtime under 1 second (Enterprise: under 60 seconds) and point-in-time recovery up to 35 days (Enterprise: up to 7 days). Both editions support MySQL, PostgreSQL and SQL Server.

How long can RDS and Cloud SQL keep backups?

RDS automated backups can be retained for 0 to 35 days, and manual snapshots are kept until you delete them. Cloud SQL automated backups can be kept for up to 10 years with enhanced backups, but point-in-time recovery is limited by log retention: up to 7 days on Enterprise and up to 35 days on Enterprise Plus.

Can the high availability standby serve reads?

Not on a Cloud SQL regional instance or an RDS Multi-AZ DB instance. An RDS Multi-AZ DB cluster (MySQL and PostgreSQL) has two reader instances that act as failover targets and serve reads. Otherwise, use read replicas.

Which is cheaper, RDS or Cloud SQL?

It depends on engine, size, HA, region, storage and commitments, and neither vendor publishes a direct comparison. Both offer discounts for one- and three-year commitments. Price the same configuration in the AWS and Google Cloud calculators.

Sources

Checked October 2026.

How we research comparisons: our editorial method.