SQL Server is the workload that determines whether an AWS migration cost case closes favourably. SQL Server licences carry the highest per-core cost of any common workload, Microsoft Software Assurance entitlement varies wildly by customer, and the AWS pricing paths (License Included, BYOL on Dedicated Hosts, RDS, Aurora PostgreSQL with Babelfish) each have non-obvious cost implications. This page is the 2026 cost reference for moving SQL Server to AWS RDS specifically, with the BYOL trap, DMS pricing, and the Babelfish alternative.
SQL Server on AWS has four common destination patterns. Each has a different cost shape and a different operational trade-off. The destination decision typically drives the migration cost case more than the source database size or schema complexity.
SQL Server on AWS, destination paths
| Destination | Licence model | Operational model | Cost vs source SQL Server |
|---|---|---|---|
| RDS for SQL Server, LI | License Included | PaaS, managed by AWS | Run-rate parity to +20% |
| RDS for SQL Server, BYOL | Bring Your Own (SA required) | PaaS, managed by AWS | -15 to -25% with SA |
| SQL Server on EC2, LI | License Included | IaaS, managed by customer | Run-rate parity, full feature set |
| SQL Server on EC2 Dedicated Hosts, BYOL | Bring Your Own (SA required) | IaaS, managed by customer | -25 to -40% with SA, but Dedicated Host minimum |
| Aurora PostgreSQL with Babelfish | No SQL Server licence required | PaaS, managed by AWS | -40 to -65% on licence; +schema conversion cost |
License Included pricing bundles the SQL Server licence cost into the hourly RDS or EC2 rate. The premium over the equivalent Linux or open-source instance is roughly $0.20 to $0.50 per hour per vCPU for SQL Server Standard, $0.80 to $1.40 per hour per vCPU for SQL Server Enterprise. BYOL on EC2 Dedicated Hosts is cheaper per vCPU but requires a minimum host commitment (typically 24 to 64 cores depending on host type), which means small SQL Server estates pay for capacity they do not use.
The break-even is roughly 8 SQL Server Standard cores or 4 Enterprise cores per Dedicated Host. Below that volume, License Included on shared tenancy RDS is the cheaper option despite the higher per-core rate. Above that volume, BYOL on Dedicated Hosts wins by 25 to 40 percent on licence cost provided the customer has Software Assurance entitlement on the full core count. The Dedicated Host minimum is the trap that catches small estates; customers with 4 SQL Server cores end up paying for 24 cores of Dedicated Host capacity.
The Software Assurance audit
AWS DMS is the primary data movement tool for SQL Server migrations to RDS. It supports full-load migration (one-time copy of the source database), full-load plus change data capture (continuous replication during the cutover window), and CDC-only (incremental updates after an initial bulk load via a different tool). DMS pricing is by replication instance class on a per-hour basis, plus storage on the replication instance and standard data transfer charges.
The current DMS pricing is on the public AWS Database Migration Service pricing page (May 2026). A representative pricing build for a 5 TB SQL Server migration with two weeks of CDC parallel replication: dms.r5.large at $0.18 per hour for 336 hours equals $60.48 of compute. Storage on the replication instance (typically 100 GB for the replication state) is $0.115 per GB-month, roughly $5.75. Data transfer in to AWS is free; data transfer out of AWS from a non-VPC endpoint is charged at standard rates. The total DMS bill for a typical 5 TB migration is around $80 to $250, dwarfed by the partner-led validation and cutover labour.
DMS Serverless was made generally available in 2023 and offers per-vCPU per-hour billing without the need to provision a replication instance class. For long-running migrations or fluctuating workloads, DMS Serverless typically saves 15 to 35 percent on tooling cost compared with provisioned DMS, at the cost of slightly higher peak-time billing during burst replication.
AWS Schema Conversion Tool (SCT) handles schema and code conversion for source-to-target combinations where the engines differ. For SQL Server to RDS for SQL Server the engine is the same, so SCT is rarely needed except for sanity-checking unsupported feature usage. For SQL Server to Aurora PostgreSQL via Babelfish, SCT is essential and typically automates 60 to 85 percent of the schema conversion, leaving the remainder for manual T-SQL to PL/pgSQL conversion.
The feature parity gaps that matter for SQL Server to RDS migrations: CLR integration is not supported on RDS, Service Broker has limited support, distributed transactions across databases require manual reconfiguration, FILESTREAM is not supported, Replication publisher mode is limited (subscriber mode is supported), Linked Servers to non-AWS targets require connectivity work, and SQL Server Reporting Services (SSRS) and Integration Services (SSIS) need to be hosted on separate EC2 instances. Each gap adds 2 to 20 hours of assessment and remediation per database depending on the feature usage.
Babelfish for Aurora PostgreSQL is a translation layer on Aurora PostgreSQL that accepts SQL Server's TDS protocol. The practical implication is that application code using SQL Server drivers can connect to Aurora PostgreSQL with Babelfish without changing the connection string or client library. The migration removes the SQL Server licence cost entirely.
The cost economics are striking but lagged. Schema conversion via SCT for SQL Server to Aurora PostgreSQL typically lands at $30K to $150K per database depending on schema complexity, T-SQL usage, and application coupling. Application-side validation adds another $20K to $80K. The full migration cost per database is materially higher than RDS for SQL Server migration cost. The run-rate saving however is roughly 40 to 65 percent per database because the SQL Server licence is removed entirely. The break-even on the migration cost versus the run-rate saving is typically 12 to 24 months for a mid-sized database.
A representative cost build for a 10-database SQL Server migration to RDS, mid-market scope, mixed sizes (3 databases under 1 TB, 5 databases 1 to 5 TB, 2 databases 5 to 20 TB), near-zero downtime requirement, 6-month programme, License Included on shared tenancy RDS for SQL Server Standard.
Worked SQL Server to AWS RDS cost build, 10 databases, 6 months
| Cost line | Low estimate | Typical estimate | High estimate |
|---|---|---|---|
| Assessment and SCT analysis | $25,000 | $50,000 | $100,000 |
| Migration labour, 10 databases, blended | $120,000 | $280,000 | $600,000 |
| AWS DMS replication instances | $2,000 | $4,500 | $10,000 |
| RDS provisioning and Multi-AZ setup | $3,000 | $6,000 | $12,000 |
| Cutover labour and validation, 10 databases | $40,000 | $90,000 | $220,000 |
| Application-side connection string updates | $15,000 | $35,000 | $80,000 |
| Reporting Services / SSIS rehosting | $20,000 | $50,000 | $120,000 |
| Parallel running, 3 months (source + RDS) | $45,000 | $95,000 | $180,000 |
| Cutover downtime contingency | $10,000 | $30,000 | $120,000 |
| Staff training (DBAs) | $8,000 | $18,000 | $32,000 |
| Security rework (IAM, encryption, KMS) | $15,000 | $35,000 | $75,000 |
| Contingency at 15 percent | $45,000 | $105,000 | $225,000 |
| Net total estimate | $348,000 | $798,500 | $1,774,000 |
The typical-column number, $800K for 10 databases over 6 months, works out at roughly $80,000 per database all-in. That is meaningfully higher than the per-database number for smaller estates because the larger databases drive the cost. Three databases of 5 to 20 TB dominate the migration labour and the parallel-running window. The cost discipline that brings the typical column toward the low column is database rationalisation before migration: many estates have legacy databases that should be retired rather than migrated.
SQL Server is the workload that most often makes or breaks a cloud migration cost case. The licence economics, the BYOL Dedicated Host trap, and the Babelfish alternative each carry meaningful financial weight. Getting the destination decision right at the assessment stage is worth more than any subsequent tooling optimisation. The teams that succeed at SQL Server migration cost are the ones that audit Software Assurance first, rationalise databases second, and only then start the cost modelling exercise.
A. A single SQL Server database under 1 TB typically costs $8,000 to $25,000 to migrate to AWS RDS for SQL Server, including DMS replication, validation, and partner-led cutover. Larger databases (5 to 50 TB) cost $25,000 to $120,000 each. The dominant variables are database size, downtime tolerance, schema complexity, and whether the source uses SQL Server features (CLR, Service Broker, Replication, FILESTREAM) that have parity gaps with RDS.
A. License Included (LI) is the simpler option and the cheaper option for most estates without active SQL Server Software Assurance. RDS for SQL Server LI bundles the SQL Server licence into the hourly rate, with no Dedicated Host requirement. BYOL on EC2 Dedicated Hosts is cheaper for estates with substantial Software Assurance entitlement, but the Dedicated Host minimums (1 host per pricing tier, typically 16+ cores) erode the saving on small estates. The break-even is roughly 8 SQL Server cores; below that, LI is usually cheaper.
A. AWS DMS uses replication instances priced like EC2. A typical DMS replication instance (dms.r5.large) costs roughly $0.18 per hour on-demand, plus storage and data transfer. For a 5 TB SQL Server migration with two weeks of parallel replication, the DMS cost lands around $600 to $1,200. The first DMS instance is free for six months under the AWS free tier promotion (current as of May 2026); larger or multiple parallel replications carry full pricing.
A. Babelfish for Aurora PostgreSQL accepts the SQL Server TDS protocol, so applications that use SQL Server client libraries can connect to Aurora without driver changes. The schema and T-SQL still need conversion via AWS Schema Conversion Tool (SCT), which automates 60 to 85 percent of typical workloads. The Babelfish path is meaningfully cheaper than RDS for SQL Server on run-rate (no SQL Server licence) but more expensive on migration cost ($30K to $150K of conversion labour per database). The break-even is typically 12 to 24 months of run-rate.
A. A single database under 1 TB completes in 2 to 4 weeks of elapsed time, with about a week of active engineering effort. Larger databases (5 to 50 TB) take 4 to 12 weeks of elapsed time. The dominant factor is downtime tolerance: customers accepting 4 to 8 hours of cutover downtime can use AWS Backup and Restore, which is cheaper and simpler. Customers requiring near-zero downtime use DMS with change data capture (CDC), which is more complex and expensive but cuts cutover to under 15 minutes.
A. RDS for SQL Server supports Multi-AZ deployments which provide automatic failover within a region. It does not directly support full SQL Server Always On Availability Groups; for that level of control customers use SQL Server on EC2 instead of RDS. The Multi-AZ deployment adds roughly 2 times the cost of a Single-AZ deployment because a synchronous standby replica is maintained. For production workloads requiring HA, Multi-AZ is the default.
AWS migration cost calculator ->
EC2, MAP credits, DataSync
AWS DMS pricing ->
Replication instance, storage, egress
Oracle to AWS ->
RDS for Oracle, Aurora PostgreSQL
On-prem to AWS ->
MAP, EC2, Snowball
AWS to Azure (RDS to SQL MI) ->
Cross-cloud SQL Server
Azure to AWS (SQL MI to RDS) ->
Cross-cloud reverse
On-prem to Azure ->
Hybrid Benefit on SQL MI
10 hidden costs ->
Detailed playbook
Strategy cost tables ->
7Rs framework
Updated 2 May 2026