We had a meeting at the office today that touched on our cloud data strategy. Much of our data currently lives in on-premises instances of Microsoft SQL Server. The conversation surfaced a question that needed answering before we can plan anything: what are our options?
After the meeting I asked Claude to help me map the full landscape. What started as a quick reference turned into a comprehensive catalog of every cloud database service I could find across the major providers — and a few independent platforms for good measure.
This post is the result. It’s long. It’s meant to be a reference, not a narrative. Bookmark it.
Amazon Web Services
AWS has the broadest database portfolio of the major cloud providers. The full lineup spans relational, NoSQL, document, graph, time-series, in-memory, and analytics services.
| Service | Management | Licensing | Migration Effort | Relative Cost | Serverless |
|---|---|---|---|---|---|
| SQL Server on EC2 | Self-managed | BYOL or included | Low | Higher | No |
| RDS for SQL Server | Fully managed | License included | Low | Medium | No |
| RDS Custom for SQL Server | Semi-managed | BYOL or included | Low | Medium | No |
| Aurora PostgreSQL | Fully managed | No license fee | Medium (with Babelfish) | Medium | Yes (v2) |
| Aurora MySQL | Fully managed | No license fee | High | Medium | Yes (v2) |
| Aurora DSQL | Serverless, distributed | No license fee | High | Varies | Yes |
| RDS for PostgreSQL | Fully managed | Open source | High | Lower | No |
| RDS for MySQL | Fully managed | Open source | High | Lower | No |
| RDS for MariaDB | Fully managed | Open source | High | Lower | No |
| RDS for Oracle | Fully managed | BYOL or included | High | Higher | No |
| RDS for Db2 | Fully managed | BYOL or marketplace | High | Higher | No |
| Redshift | Fully managed warehouse | No license fee | High (analytics) | Medium-Higher | Yes |
| DynamoDB | Serverless NoSQL | No license fee | High (re-architecture) | Lower | Yes |
| Neptune | Fully managed graph | No license fee | High | Medium-Higher | Yes |
| DocumentDB | Fully managed document | No license fee | High | Medium | Yes |
| Keyspaces | Serverless wide-column | No license fee | High | Lower | Yes |
| Timestream | Fully managed time-series | Open source (InfluxDB) | High | Medium | No |
| ElastiCache | Fully managed cache | Open source | N/A | Lower | Yes |
| MemoryDB | Fully managed in-memory | Open source | N/A | Medium-Higher | No |
| OpenSearch Service | Fully managed search | Open source | N/A | Medium-Higher | Yes |
ElastiCache, MemoryDB, and OpenSearch are complementary services — they’re not SQL Server replacements but are commonly deployed alongside a primary database for caching, durable in-memory workloads, and search, respectively. Aurora PostgreSQL with Babelfish deserves special mention: it accepts T-SQL and the TDS wire protocol, which can significantly reduce migration effort for SQL Server applications. Aurora DSQL is the newest entrant, a fully serverless distributed SQL database with active-active multi-region support that went GA in May 2025.
A word on DynamoDB. The table above marks it as “High” migration effort from SQL Server, and that’s accurate — you can’t take a relational schema and drop it into a NoSQL key-value store. But DynamoDB isn’t trying to replace SQL Server. It shines in a completely different set of use cases: web and mobile backends with predictable access patterns, session management, user profiles, real-time leaderboards, IoT telemetry, and any workload where you need single-digit-millisecond reads at virtually any scale. I use DynamoDB for all of my personal web projects. It’s serverless, scales to zero cost when idle, handles traffic spikes without provisioning, and pairs naturally with Lambda and API Gateway. If you’re building something new on AWS rather than migrating something old, it’s worth serious consideration — just don’t try to force a relational workload into it.
Microsoft Azure
For SQL Server shops, Azure is the most natural starting point — same vendor, same engine. Azure offers the broadest set of SQL Server-compatible managed services, including Azure SQL Managed Instance, which provides roughly 95% compatibility with on-premises SQL Server — including SQL Agent, CLR, cross-database queries, and linked servers.
| Service | Management | Licensing | Migration Effort | Relative Cost | Serverless |
|---|---|---|---|---|---|
| SQL Server on Azure VM | Self-managed | BYOL or included | Low | Varies | No |
| Azure SQL Database | Fully managed | BYOL (AHB) or included | Medium | Medium | Yes |
| Azure SQL Elastic Pool | Fully managed | BYOL (AHB) or included | Medium | Lower-Medium | No |
| Azure SQL Managed Instance | Fully managed | BYOL (AHB) or included | Low | Medium-Higher | No |
| Azure DB for MySQL | Fully managed | Open source | High | Lower | No |
| Azure DB for PostgreSQL | Fully managed | Open source | High | Lower | No |
| Cosmos DB for NoSQL | Fully managed document | No license fee | High | Varies | Yes |
| Cosmos DB for MongoDB | Fully managed document | No license fee | High | Varies | Yes |
| Cosmos DB for Cassandra | Fully managed wide-column | No license fee | High | Varies | Yes |
| Cosmos DB for Gremlin | Fully managed graph | No license fee | High | Varies | Yes |
| Cosmos DB for Table | Fully managed key-value | No license fee | High | Medium | Yes |
| Cosmos DB for PostgreSQL | Distributed relational | Open source (Citus) | High | Medium | No |
| Azure Synapse Analytics | Fully managed warehouse | No license fee | Medium-High (analytics) | Varies | Yes (serverless pools) |
| Azure Data Explorer | Fully managed analytics | No license fee | High | Medium-Higher | No |
| Azure Managed Redis | Fully managed cache | No license fee | N/A | Medium | No |
| Azure Table Storage | Fully managed key-value | No license fee | High | Lower | Yes |
| Azure Confidential Ledger | Fully managed ledger | No license fee | High | Lower | Yes |
| Managed Instance for Cassandra | Managed instance | Open source | High | Medium | No |
Azure Hybrid Benefit (AHB) lets you apply existing SQL Server licenses with Software Assurance to Azure SQL services, saving up to 55%. Combined with reserved capacity, savings can reach 80%. Azure SQL Database offers a serverless compute tier that auto-pauses when idle — you pay nothing for compute during idle periods. Azure Database for MariaDB was retired in September 2025. Azure Cache for Redis is being retired in favor of Azure Managed Redis. On the horizon: Azure HorizonDB, a PostgreSQL-compatible cloud-native database in private preview, and Microsoft Fabric Databases, a unified SaaS data platform that went GA in November 2025.
Google Cloud Platform
Google Cloud has a smaller database portfolio than AWS or Azure, but it includes some notably differentiated cloud-native options — particularly Cloud Spanner and AlloyDB.
| Service | Management | Licensing | Migration Effort | Relative Cost | Serverless |
|---|---|---|---|---|---|
| SQL Server on Compute Engine | Self-managed | BYOL or included | Low | Varies | No |
| Cloud SQL for SQL Server | Fully managed | License included | Low | Medium | No |
| Cloud SQL for MySQL | Fully managed | Open source | High | Lower | No |
| Cloud SQL for PostgreSQL | Fully managed | Open source | High | Lower | No |
| AlloyDB for PostgreSQL | Fully managed, cloud-native | Proprietary (PG-compatible) | High | Medium-Higher | No |
| Cloud Spanner | Fully managed, global | Proprietary | High | Higher | Yes (autoscaler) |
| BigQuery | Serverless warehouse | Proprietary | High (analytics) | Lower-Varies | Yes |
| Cloud Bigtable | Fully managed wide-column | Proprietary | High | Higher (small scale) | Autoscaling |
| Firestore | Serverless document | Proprietary | High | Lower | Yes |
| Firebase Realtime Database | Serverless JSON | Proprietary | High | Lower | Yes |
| Memorystore for Redis | Fully managed cache | Open source | N/A | Medium | No |
| Memorystore for Valkey | Fully managed cache | Open source (BSD) | N/A | Medium | No |
AlloyDB is Google’s cloud-native PostgreSQL-compatible engine, claiming 4x faster transactions and up to 100x faster analytics than standard PostgreSQL. It also offers AlloyDB Omni, a downloadable version that runs on-premises, in containers, or on other clouds. Cloud Spanner is Google’s globally distributed, strongly consistent relational database supporting both GoogleSQL and PostgreSQL dialects, with recently added graph, full-text search, and vector search capabilities. It offers a managed autoscaler and tiered storage (SSD + HDD) to manage costs. Google’s Database Migration Service is free for same-engine migrations and now supports heterogeneous SQL Server-to-PostgreSQL migrations with Gemini-assisted conversion.
Oracle Cloud and IBM Cloud
Oracle and IBM have smaller cloud database portfolios but include some differentiated offerings.
Oracle Cloud Infrastructure
| Service | Management | Licensing | Migration Effort | Relative Cost | Serverless |
|---|---|---|---|---|---|
| Autonomous Database | Self-driving, fully managed | BYOL or included | High | Medium-Higher | Yes |
| Base Database Service | Infra-managed, DBA-controlled | BYOL or included | High | Medium | No |
| MySQL HeatWave | Fully managed | Open source | Medium-High | Lower-Medium | No |
| OCI PostgreSQL | Fully managed | Open source | Medium-High | Lower | No |
| NoSQL Database Cloud | Fully managed | No license fee | Very High | Lower | Yes (on-demand) |
Oracle’s Autonomous Database is genuinely self-managing — it handles patching, tuning, and scaling without DBA intervention. MySQL HeatWave includes an integrated in-memory analytics engine and is available on OCI, AWS, and Azure. Oracle also runs natively within Azure and Google Cloud datacenters through multi-cloud partnerships.
IBM Cloud
| Service | Management | Licensing | Migration Effort | Relative Cost | Serverless |
|---|---|---|---|---|---|
| Db2 on Cloud | Fully managed | Proprietary | Medium | Medium | No |
| Cloud Databases for PostgreSQL | Fully managed | Open source | Medium-High | Medium | No |
| Cloud Databases for MongoDB | Fully managed | Proprietary | Very High | Medium-High | No |
| Cloud Databases for EDB | Fully managed | EDB license | Medium-High | Medium-High | No |
IBM Db2 is one of the closer dialect matches to SQL Server among non-Microsoft relational databases, which may reduce stored procedure conversion effort compared to PostgreSQL or MySQL targets.
Multi-Cloud and Independent Platforms
Several database platforms run across multiple cloud providers, avoiding lock-in to any single vendor.
| Service | Type | Licensing | Migration Effort | Relative Cost | Serverless | Runs On |
|---|---|---|---|---|---|---|
| Snowflake | Data warehouse | Proprietary | Medium (analytics) | Medium | Yes | AWS, Azure, GCP |
| MongoDB Atlas | Document NoSQL | Proprietary | Very High | Lower-Medium | Yes (Flex tier) | AWS, Azure, GCP |
| CockroachDB Cloud | Distributed SQL (PG-compat) | Proprietary | Medium-High | Medium | Yes (Basic tier) | AWS, Azure, GCP |
| Neon | Serverless PostgreSQL | Open source | Medium-High | Very Low | Yes | AWS |
| PlanetScale | MySQL/PG (Vitess) | Proprietary | Medium-High | Lower-Medium | No | AWS, GCP |
| TiDB Cloud | Distributed SQL (MySQL-compat) | Open source | Medium-High | Lower-Medium | Yes (Starter) | AWS, GCP |
| YugabyteDB Aeon | Distributed SQL (PG-compat) | Proprietary | Medium-High | Medium-Higher | No | AWS, GCP |
| Aiven | Multi-engine platform | Open source | Varies | Lower-Medium | No | AWS, Azure, GCP+ |
| Crunchy Bridge | Managed PostgreSQL | Open source | Medium-High | Lower | No | AWS, Azure, GCP |
| Supabase | PostgreSQL + backend | Open source | Medium-High | Lower | Partial | AWS |
Snowflake is specifically an analytics platform, not an OLTP database — it includes SnowConvert AI for automated T-SQL translation and a dedicated SQL Server migration guide. Neon offers true scale-to-zero serverless PostgreSQL with a dedicated SQL Server migration guide. CockroachDB and YugabyteDB are distributed SQL databases designed for global, multi-region deployments with strong consistency. MongoDB Atlas provides a Relational Migrator that can help convert SQL Server schemas to document models.
Cloud-Native and Consumption-Based Options
Some of these services were built from the ground up for the cloud — no on-premises ancestor, no legacy architecture. Several offer consumption-based pricing where you pay only for what you use rather than provisioning fixed capacity. Here’s which ones meet that description:
| Service | Provider | Type | Billing Model |
|---|---|---|---|
| Aurora Serverless v2 | AWS | Relational (PostgreSQL/MySQL) | Per ACU-hour, per second |
| Aurora DSQL | AWS | Distributed SQL | Per DPU + storage |
| DynamoDB | AWS | NoSQL key-value/document | Per request + storage |
| Redshift Serverless | AWS | Data warehouse | Per RPU-hour, per second |
| Neptune Serverless | AWS | Graph | Per NCU-hour, per second |
| DocumentDB Serverless | AWS | Document (MongoDB-compatible) | Per DCU-hour, per second |
| Keyspaces | AWS | Wide-column (Cassandra-compatible) | Per request + storage |
| ElastiCache Serverless | AWS | In-memory cache | Per ECPU + GB-hours |
| Azure SQL Database | Azure | Relational (SQL Server) | Per vCore-second, auto-pause |
| Cosmos DB | Azure | Multi-model NoSQL | Per RU consumed + storage |
| Synapse Serverless Pools | Azure | Data warehouse | Per TB processed |
| Azure Table Storage | Azure | Key-value | Per transaction + storage |
| BigQuery | Data warehouse | Per TiB scanned + storage | |
| Firestore | Document NoSQL | Per operation + storage | |
| Cloud Spanner | Global relational | Per processing unit + storage | |
| Oracle Autonomous DB | Oracle | Relational (Oracle) | Per ECPU-hour + storage |
| Snowflake | Multi-cloud | Data warehouse | Per credit consumed + storage |
| Neon | Multi-cloud | Serverless PostgreSQL | Per CU-hour + storage |
| CockroachDB Basic | Multi-cloud | Distributed SQL | Per Request Unit + storage |
| MongoDB Atlas Flex | Multi-cloud | Document NoSQL | Per operation (capped) |
| TiDB Cloud Starter | Multi-cloud | Distributed SQL | Per Request Unit + storage |
What I Took Away
The landscape is enormous. What started as a straightforward question — “should we move SQL Server to the cloud?” — turns out to have dozens of possible answers depending on what you optimize for.
For organizations like mine that run on SQL Server today, the lowest-friction paths are clear. Lift-and-shift to a VM, or step up to a managed service that runs the same engine. On AWS, that’s RDS for SQL Server or RDS Custom. On Azure, it’s Azure SQL Managed Instance — probably the single easiest managed migration path because Microsoft built it specifically for on-prem SQL Server workloads, with roughly 95% feature compatibility. On Google Cloud, it’s Cloud SQL for SQL Server.
But “lowest friction” and “best long-term fit” aren’t always the same thing. The cloud-native and consumption-based services are architecturally different — many of them eliminate database administration entirely and scale automatically with demand. The tradeoff is migration effort. Moving from SQL Server to DynamoDB or Cosmos DB isn’t a migration — it’s a re-architecture. Moving to Aurora PostgreSQL with Babelfish sits somewhere in between: you get a cloud-native engine with T-SQL compatibility that reduces — but doesn’t eliminate — the conversion work.
And sometimes the answer isn’t migrating at all — it’s building the next thing differently. DynamoDB isn’t going to replace our SQL Server estate, but when I build a new web project from scratch, it’s where I start. No servers, no patching, no capacity planning. The right database depends on the workload, not on what you’ve always used.
I’m not making recommendations here. This is a map, not a GPS route. But having the full picture in one place makes the next conversation a lot more productive.
Sources
- AWS Database Services Overview
- Amazon Aurora Babelfish
- Amazon Aurora DSQL
- Azure SQL Documentation
- Azure SQL Managed Instance Migration Guide
- Azure Hybrid Benefit
- Azure Database for MariaDB Retirement
- Azure HorizonDB
- Microsoft Fabric Databases GA
- Google Cloud Databases
- Google AlloyDB
- Google Cloud Spanner
- Google Database Migration Service
- Oracle Autonomous Database
- Oracle MySQL HeatWave
- IBM Db2 on Cloud
- Snowflake SQL Server Migration Guide
- MongoDB Atlas Relational Migrator
- CockroachDB Migration Overview
- Neon SQL Server Migration Guide