Migrating to AWS: the Cloud Adoption Framework, the 7 Rs and the tools
Everything in AWS I-VI assumed the workload was already in AWS. Most companies start with hundreds of applications in their own data centre, and the first architecture question is not "Lambda or Fargate?" but "what do we do with each of these, in which order, and why at all?" This lesson is the map for that: why companies migrate, how AWS structures the organisational side (the Cloud Adoption Framework), the seven migration strategies you pick per application, and the services that move servers, databases and files.
Need to know: a migration moves workloads from somewhere else (a data centre, another cloud) to AWS; modernization changes how they are built (managed services, containers, serverless) and can happen during the move or after it. You decide per application, not for the whole company. (simulator) The migration services themselves - AWS Transform, Application Migration Service, DMS, DataSync, Transfer Family - are not simulated; the commands in this lesson read a portfolio inventory and config files on the box, the way you prepare a migration plan before any of them runs.
Why migrate: the business outcomes
A migration costs money and attention, so it needs a business case. AWS names four business outcomes that cloud adoption improves - the Cloud Practitioner exam asks for them by name:
| Outcome | What it means in practice |
|---|---|
| Reduced business risk | better reliability and security than the old data centre: Multi-AZ, backups, patched managed services, no end-of-life hardware |
| Improved environmental, social and governance (ESG) performance | AWS's data centres use less energy per workload than a typical company's; sustainability reporting |
| Increased revenue | new products and features ship faster; new markets and Regions in days, not a data-centre project |
| Increased operational efficiency | less time on racking, patching and capacity planning; pay for what you use instead of buying for the peak |
The money side is what AWS III called cloud economics: capital expense (servers bought up front, sized for the peak, written off over years) becomes operating expense that follows usage. A total cost of ownership (TCO) comparison has to include what the data centre really costs - power, cooling, space, hardware refreshes, licences and the people who run it - not only the server price. Migration Evaluator (a free AWS service) builds such a business case from measured utilisation of the current servers.
The AWS Cloud Adoption Framework (AWS CAF)
AWS CAF is AWS's guidance for the organisational side of moving to the cloud: which capabilities a company needs to build, and who owns them. It groups them into six perspectives, three business and three technical:
| Perspective | Focus | Typical owners | Example capabilities |
|---|---|---|---|
| Business | cloud investments that serve the business strategy | CEO, CFO, COO, CIO | strategy, portfolio, product and data management |
| People | culture, skills and organisational change | CHRO, HR, leaders | cloud fluency, workforce transformation, organisation design |
| Governance | getting the benefits while managing risk | CFO, CIO, program managers | program and project management, risk, cloud financial management |
| Platform | a scalable, hybrid cloud platform | CTO, architects, engineers | architecture, platform engineering, deployment automation, data engineering |
| Security | confidentiality, integrity and availability of data and workloads | CISO, security engineers | identity, threat detection, incident response, data protection |
| Operations | cloud services delivered at the agreed level | IT operations, SRE | observability, event and incident management, patching, availability |
Memory hook: Business, People, Governance = business capabilities; Platform, Security, Operations = technical capabilities. A stalled migration is usually a business-side problem - people afraid for their jobs (People), no budget owner (Governance) - not a missing service.
CAF also describes the journey in four iterative phases: Envision (which business outcomes, measured how), Align (find the capability gaps across the six perspectives and the dependencies between teams), Launch (pilot initiatives in production that show value) and Scale (expand the pilots to the rest of the business). The change it aims at happens in four transformation domains: technology, process, organization and product.
For the migration project itself, AWS uses three phases: assess (inventory, business case, readiness), mobilize (landing zone - accounts, network, identity, as in AWS IV - plus skills, wave planning, and a first few applications), then migrate and modernize in waves.
The 7 Rs: one strategy per application
AWS Prescriptive Guidance names seven migration strategies. Sorting the portfolio into them is the core of the assess phase:
| Strategy | What happens | When | Effort |
|---|---|---|---|
| Retire | switch it off | nobody uses it, or another system already does the job | none, and it saves money |
| Retain | keep it where it is, for now | recently upgraded hardware, a compliance or latency reason, a mainframe not ready to move, no business value in moving | none now |
| Rehost ("lift and shift") | move the servers as they are to Amazon EC2 | many applications, a deadline (the data-centre lease ends), no changes allowed | low |
| Relocate | move a whole platform without changing it: VMware VMs to VMware Cloud on AWS or Amazon Elastic VMware Service, containers between clusters | large VMware estates, keep the operating model | low |
| Replatform ("lift, tinker and shift") | move with one or two changes for a clear benefit, without changing the architecture: the database to Amazon RDS, the app to Elastic Beanstalk or containers | reduce administration (patching, backups) quickly | medium |
| Repurchase ("drop and shop") | replace the product with another one, usually SaaS: the self-hosted CRM becomes a SaaS CRM | a commodity application, a licence model that does not fit the cloud | medium (data and process migration) |
| Refactor / re-architect | rebuild it cloud-native: microservices, serverless, managed data stores, as this chapter's first lessons did | the application must scale, change faster, or cannot meet a requirement as it is | high, and the highest long-term benefit |
Typical order: retire and retain first (they shrink the project), rehost or relocate most of the rest quickly, replatform where one change pays off at once, refactor the few applications where the business needs it - often after they are already in AWS ("migrate first, then modernize").
Sort a portfolio
The try- migration has a portfolio inventory exported from the old data centre's CMDB, one line per application:
$ cd ~/oncall-lab/labs/aws/try6
$ column -t -s, portfolio.csv
app owner os database users_90d note strategy
orders-web shop Ubuntu 20.04 - 4210 no changes before the lease ends rehost
orders-api shop Ubuntu 20.04 PostgreSQL 13 4210 must scale for sales events refactor
billing finance RHEL 8 PostgreSQL 13 312 DBAs want out of patching replatform
crm sales Windows 2016 SQL Server 2016 95 vendor sells it as SaaS repurchase
fax-gateway ops CentOS 7 - 0 last fax in 2023 retire
old-intranet it Windows 2012 - 0 replaced by the wiki retire
plant-control factory Windows 2019 - 40 must stay next to the machines retain
vm-farm it vSphere 8 - 820 140 VMs; keep the VMware tooling relocate
reporting finance Ubuntu 22.04 MySQL 8 58 nightly batch jobs rehost
hr-portal hr RHEL 8 Oracle 19c 600 Oracle licence ends next year replatform
Read it like an assessment: users_90d = distinct users in the last 90 days, note = what the owners said. The two zero-user applications are the first decision of every migration - and the cheapest:
Then by strategy, the way a wave plan starts:
$ awk -F, 'NR > 1 { print $7 }' portfolio.csv | sort | uniq -c | sort -rn
2 retire
2 replatform
2 rehost
1 retain
1 repurchase
1 relocate
1 refactor
Replatform in one diff
billing is marked replatform: its self-managed PostgreSQL moves to Amazon RDS for PostgreSQL, the application stays as it is. The change you review is a config file, and diff shows exactly what the migration changes:
The full picture, in the format a pull request shows it:
$ diff -u billing-onprem.conf billing-aws.conf | tail -n +3
@@ -1,5 +1,5 @@
# billing service
-db.host=pg-billing-01.dc1.internal
+db.host=billing.c9k2m4xq7v1d.eu-central-1.rds.amazonaws.com
db.port=5432
db.name=billing
db.user=billing_app
The tools
| Need | Service | What it does |
|---|---|---|
| discover, assess, plan waves, modernize | AWS Transform | AWS's migration and modernization service (launched May 2025): discovery and assessment, wave planning, rehost of VMware workloads, and AI agents that modernize .NET, mainframe and VMware applications. It replaces AWS Migration Hub and AWS Application Discovery Service, which closed to new customers on 7 November 2025 |
| a business case | Migration Evaluator | measured utilisation and licences -> projected AWS cost |
| servers (rehost) | AWS Application Migration Service (MGN) | installs an agent on the source servers, replicates their disks continuously at block level into AWS, launches test instances, then cuts over with minutes of downtime |
| databases | AWS Database Migration Service (DMS) | copies the data (full load) and then keeps replicating changes (change data capture), so the source stays live until cutover; same engine (homogeneous) or a different one (heterogeneous) |
| a different database engine | DMS Schema Conversion / AWS Schema Conversion Tool (SCT) | converts the schema and code (views, procedures) between engines, for example Oracle to Aurora PostgreSQL; DMS then moves the data |
| files and objects, online | AWS DataSync | transfers from NFS, SMB, HDFS and object storage to S3, EFS or FSx, encrypted and verified, scheduled or one-off; also between AWS storage services |
| partners who send files | AWS Transfer Family | managed SFTP, FTPS, FTP and AS2 endpoints that store into S3 or EFS, so partners keep their tools |
| a lot of data, offline | Snow Family (legacy) | Snowball Edge devices shipped to you, filled, shipped back. Closed to new customers since 7 November 2025 (Snowcone and Snowmobile were retired earlier); AWS now points to DataSync over Direct Connect, the AWS Data Transfer Terminal (sites where you bring your own storage to upload at high speed) or partners |
"Database replication", the Cloud Practitioner exam's example strategy, is the DMS pattern above, or the engine's own replication (a PostgreSQL streaming replica, as in the PostgreSQL chapter, or an RDS read replica from an external MySQL source): the new database follows the old one until the moment you switch the application over, so the downtime is the switch, not the copy.
Online or offline is arithmetic. A link moves its speed in bits per second; storage is in bytes:
On the exam (CLF-C02 1.3 "benefits of and strategies for migration"; SAA-C03 3.5 for the transfer services): the four CAF business outcomes and the six perspectives (business vs technical); the 7 Rs from a scenario ("no changes, fast" = rehost, "new connection string to RDS" = replatform, "SaaS instead" = repurchase, "nobody uses it" = retire, "must stay on premises for now" = retain); Application Migration Service for servers, DMS (+ Schema Conversion) for databases with the source live, DataSync for files online, Transfer Family for SFTP partners. Answers built on AWS Migration Hub, Application Discovery Service or Snowball Edge are out of date for new customers.
In an interview: "How would you plan the migration of 300 applications?" - "Assess first: an inventory with owners, usage and dependencies, and a business case. Retire what nobody uses, retain what cannot move yet, rehost most of the rest in waves with Application Migration Service, replatform databases to RDS with DMS where it pays off at once, and refactor only the few applications that need it, usually after they are in AWS. Before the first wave: a landing zone, skills, and a pilot that shows the value."
You can now: name the four business outcomes AWS CAF promises and its six perspectives, split into business and technical, describe the envision-align-launch-scale phases, pick one of the 7 Rs for an application from its description, sort a portfolio inventory with awk, show a replatform's change with diff, and choose the AWS service that moves servers, databases, files or partner transfers - knowing which older names are closed to new customers.