4.2 - Executing migrations (MGN, DMS, Snowball, 7 Rs strategies)
AWS Solutions Architect Professional objective 4.2 covers executing a migration and running cutovers. You rehost Linux VMs with the least effort using AWS Application Migration Service (MGN), which continuously replicates source servers at the block level into a low-cost staging area and launches non-disruptive test instances before the final cutover, and you replatform to Amazon RDS for MySQL to stop patching and managing failover by hand. AWS Database Migration Service needs change data capture (CDC) enabled so writes during and after the full load reach the target, and a homogeneous Oracle-to-RDS move needs no Schema Conversion Tool while a heterogeneous Oracle-to-Aurora PostgreSQL move does. Snowball Edge ships offline bulk data and runs local compute, a Tape Gateway retires physical tape to Glacier, and a repurchase swaps a customized CRM for SaaS. For a governed cutover you keep source servers running until validation succeeds and schedule a low-traffic window. Expect scenarios about a lift-and-shift, missing post-load rows, or a safe cutover, and ask which tool or strategy fits.
Lift-and-shift servers with least effort = rehost with MGN (block-level replication + test instances). Target missing rows written after the full load = DMS change data capture (CDC) was off. Same engine = no SCT; different engine = SCT needed.
Practice questions
1. A company must migrate a 50 TB on-premises dataset to Amazon S3 within a week, but its internet link would take over a month to transfer that volume. What is the best approach?
- Compress the data and upload it over the existing internet link
- Order an AWS Snowball Edge device, copy the data, and ship it back (correct answer)
- Set up a site-to-site VPN and copy the data with the AWS CLI
- Use S3 Transfer Acceleration from the on-premises servers
When the network would take too long for a large one-time transfer, Snowball Edge physically ships the data and lands it in S3 within days. Compression, VPN, and Transfer Acceleration all still ride the constrained link and cannot beat the deadline.
2. A team must migrate a production Oracle database to Aurora PostgreSQL with minimal downtime, keeping the source live until cutover. Which tools fit best?
- AWS SCT to convert the schema, then AWS DMS with ongoing replication (CDC) (correct answer)
- A one-time mysqldump exported and imported during a maintenance window
- S3 export of Oracle files restored directly into Aurora
- EC2-hosted Oracle with a read replica in Aurora
For a heterogeneous engine change (Oracle to PostgreSQL), the Schema Conversion Tool converts schema/code and DMS with change data capture replicates ongoing changes so the source stays live until a short cutover. A dump needs long downtime; you cannot restore Oracle files into Aurora, nor create a cross-engine read replica this way.
3. A company must move 40 Linux VMs to AWS within a tight deadline and cannot afford to change the applications' code. Which of the 7 Rs strategies fits, with the least engineering effort?
- Refactor each application into event-driven microservices before starting the move
- Replatform by swapping every database engine during the migration
- Rehost (lift-and-shift) the servers with no application changes (correct answer)
- Repurchase by rebuilding on a third-party SaaS equivalent product
Rehost moves servers as-is with no code change, so it is fastest and lowest effort for a tight deadline. Refactor and replatform both require engineering work, and repurchase means abandoning the current app for a SaaS product, which is a larger business change.
4. A team wants to rehost hundreds of physical and virtual servers to EC2 with continuous block-level replication, non-disruptive testing, and a short cutover window. Which service automates this at scale?
- AWS Application Migration Service (MGN) (correct answer)
- AWS Backup copying volumes on a schedule
- Amazon EC2 Image Builder producing golden AMIs
- AWS Elastic Disaster Recovery for failover only
MGN continuously replicates source disks to a staging area, lets you launch test instances without disrupting the source, and performs a quick cutover, making it AWS's primary rehost tool at scale. Backup and Image Builder do not orchestrate live migration, and Elastic DR targets disaster recovery rather than migration.
5. An aging self-hosted CRM is costly to maintain and offers no strategic value. The business is willing to adopt a commercial cloud CRM and re-import its data. Which R strategy is this?
- Rehost the existing CRM servers unchanged onto EC2 instances
- Refactor the CRM into event-driven microservices on Lambda
- Retain the CRM on-premises and revisit the decision later
- Repurchase: replace it with a SaaS CRM (correct answer)
Repurchase (drop-and-shop) means abandoning the custom app and moving to a different, usually SaaS, product, which fits a commodity CRM with no strategic value. Rehost and refactor keep the same application, and retain leaves it running on-premises rather than replacing it.
6. A remote site has 60 TB to move to AWS but only a slow, unreliable internet link, so an online copy would take months. Which TWO facts make AWS Snowball Edge the right choice? (Choose TWO.)
- Data is shipped physically, avoiding the constrained network entirely (correct answer)
- It requires provisioning a dedicated 10 Gbps Direct Connect circuit before any transfer can begin
- It ships a rugged device and encrypts data at rest during transit (correct answer)
- It streams data continuously and needs no local staging capacity
Snowball Edge sidesteps the slow link by moving data on a physical, encrypted device you ship to AWS, which is faster than months of online transfer. It does not require Direct Connect, and it is offline transport, not continuous streaming.