Asterrr's Handbook

Assessment and the 7 Rs

Picking a migration strategy per application, building the business case, discovering servers and dependencies, and planning waves.

Exam tasks: 4.1 (select existing workloads and processes for potential migration, determine the optimal migration approach, evaluate application migration tooling)

The decision: for each application in the portfolio, which of the 7 Rs fits the business goal and deadline, and what data do you need to collect before you can say so with confidence?

The migration phases

AWS frames a large migration as three phases. Questions often describe an activity and expect you to know which phase it belongs to, or which tool supports it.

PhaseWhat happensTypical tools and outputs
AssessDecide whether to migrate and why. Build a directional business caseMigration Evaluator (TCO), Migration Readiness Assessment, high-level inventory
MobilizeClose the gaps the assessment found. Detailed discovery, dependency mapping, a 7R decision per appApplication Discovery Service, landing zone with Control Tower, a cloud center of excellence, pilot waves
Migrate and modernizeMove applications in waves, validate, cut over, then optimizeApplication Migration Service, DMS, DataSync, wave runbooks

Landing zone first

If a question says the company is about to migrate hundreds of servers into AWS and has no account structure, the right first step is a multi-account landing zone (Control Tower), not starting replication. That belongs in Mobilize. See multi-account governance.

The 7 Rs

Least effort · least changeMost effort · most cloud benefit
  1. Retire
    Turn it off

    The application has no users or is duplicated elsewhere. Decommission it and archive the data.

  2. Retain
    Leave it where it is

    Not ready, recently upgraded, tied to hardware or a compliance boundary, or not worth moving yet. Revisit later.

  3. Rehost
    Lift and shift

    Move servers as they are onto EC2, usually with Application Migration Service. Fastest path at scale.

  4. Relocate
    Move the hypervisor

    Move VMs to a cloud version of the same platform, such as VMware Cloud Foundation on Amazon EVS. No change to the VMs.

  5. Repurchase
    Drop and shop

    Replace the application with a SaaS product or a newer licensed version, for example a self-hosted CRM with a SaaS CRM.

  6. Replatform
    Lift, tinker and shift

    Keep the core code but swap a component for a managed service, such as a self-managed database for RDS.

  7. Refactor
    Re-architect

    Rewrite for cloud-native features: microservices, containers, serverless, purpose-built databases.

A decision tree per application

StrategyTime to migrateCode changeOngoing ops benefitExam signal
RetireNoneNoneRemoves cost"Nobody has logged in for a year"
RetainNoneNoneNone yet"Contract ends in 18 months", "mainframe not in scope"
RehostFastNoneLow at first"Data center lease ends in 4 months", "hundreds of servers"
RelocateFastest for VMware estatesNoneLow"Keep vSphere tooling and skills", "no VM conversion"
RepurchaseVariesReplacedHigh"Commodity function", "licensing model changing"
ReplatformMediumMinimalMedium"Reduce DBA effort", "managed patching and backups"
RefactorSlowSignificantHighest"Scale to 10x", "release weekly", "remove the monolith"

Refactoring under a hard deadline

When a question pairs a tight exit date with a large estate, answers that refactor everything first are wrong. The usual pattern is rehost now to meet the deadline, then replatform or refactor once in AWS. Modernizing after the move is cheaper because you're no longer paying for two data centers.

Rehost is not 'no change at all'

Rehosted servers still need new IP addresses or DNS, security groups, backup policies and monitoring. Rehost means the application code and OS stay the same, not that operations stay the same.

Building the business case: Migration Evaluator

Migration Evaluator produces a directional TCO and a projected AWS cost for the discovered estate.

  • It deploys a collector on premises that gathers utilization over a period of weeks, or it can use existing inventory exports.
  • It right-sizes to EC2 based on actual use rather than provisioned size, and models licensing options such as bring your own license versus license included for Windows and SQL Server.
  • Output is a business case report for executives. It's an Assess-phase tool, not a migration tool.

Discovery: Application Discovery Service

AWS Application Discovery Service (ADS) collects server inventory, utilization and dependencies, and stores it in the Migration Hub home Region.

Agentless CollectorDiscovery Agent
Deployed asOne OVA virtual appliance in VMware vCenterSoftware installed on every server
Works forVMware VMs onlyPhysical servers, any hypervisor, VMs in other clouds, Windows and Linux
CollectsVM inventory, configuration, CPU, memory and disk utilization from vCenterAll of that, plus running processes and inbound and outbound network connections
Dependency mappingNot from the VM layerYes: which servers talk to which, on which ports
EffortLow: one appliance per vCenterHigher: agent rollout through config management
Database inventoryOptional database and analytics module for supported enginesNo
  • A CSV import is the fallback when you can't install anything. You get inventory but no utilization or dependency data.
  • Agent data can be exported to S3 and queried with Athena for custom dependency analysis.

Dependencies mean agents

Whenever the question needs network dependencies between servers to group them into waves, or the estate isn't all VMware, the answer is the Discovery Agent. For a quick, low-touch inventory of a VMware estate, it's the Agentless Collector.

Tracking: Migration Hub

AWS Migration Hub is the single place to see discovered servers and track migrations across tools.

  • Home Region. All discovery and tracking data lives in one Region you choose, regardless of where you migrate.
  • Tracking. Application Migration Service and DMS report status into Migration Hub, grouped by application.
  • Strategy Recommendations. Analyzes discovered applications, including source code and binaries through a collector, and suggests a 7R path and target service such as containers or managed databases.
  • EC2 instance recommendations from discovered utilization, with purchasing options.
  • Orchestrator runs templated migration workflows, and Journeys guides teams through the phases.

2025 availability change

AWS Migration Hub (including Strategy Recommendations, Orchestrator, Journeys and Refactor Spaces) and Application Discovery Service stopped accepting new customers on November 7, 2025. Existing customers can keep using them, and AWS Transform is the recommended successor for new projects. SAP-C02 still tests the concepts above, so learn what each one does.

Wave planning

A wave is a group of applications that migrate and cut over together.

Group by dependency. Servers that talk to each other constantly, such as an app tier and its database, go in the same wave so you don't create cross-site latency during the migration.

Start small. Pilot waves use low-risk, low-dependency applications to prove the tooling, runbooks and network.

Grow the waves. Later waves get bigger as the factory matures. Business-critical apps move once patterns are proven.

Plan cutover windows. Match each wave to a business-approved downtime window, with rollback criteria.

Scenarios

Scenario
A logistics company must vacate its data center in 5 months. It runs 900 servers: 700 VMware VMs and 200 physical Linux servers. The architect needs utilization data and a map of which servers communicate, to group servers into migration waves. Which approach meets the requirements?
Scenario · choose 2
A retailer is reviewing its application portfolio. An internal HR tool has had no logins for 14 months. A payroll system is being replaced by a SaaS product next year. An order service must handle 10x holiday traffic, which its current design can't do. Which TWO strategy assignments are correct?
Scenario
A CFO wants a cost comparison between the current on-premises estate and AWS before approving a migration program. Few servers have been inventoried. What should the architect use first?

Further reading

On this page