Migration Strategy, Assessment, and Planning
What this domain tests
This domain covers building the case and the plan: inventorying the estate, choosing a migration pattern per workload, sizing targets, sequencing waves, and defining success measures. Questions test whether you pick the right OCI tool for discovery and assessment and can justify the pattern you chose.
Core rules and concepts
- Classify every workload with the 6 Rs: rehost (lift and shift), replatform (minor change, e.g., to Autonomous Database), refactor/re-architect, repurchase (SaaS), retire, retain.
- Discovery precedes design. OCI Cloud Migrations discovery collects server inventory, installed software, network dependencies, and utilization; dependency mapping drives wave grouping and prevents orphaned tiers.
- Assessment output feeds two decisions: target shape/size and migration method (agent replication vs export/import vs rebuild). Size from observed peaks plus agreed headroom, not from on-prem hardware nameplates.
- Oracle Cloud Lift Services and Oracle Migration Factory provide funded guidance and factory-style execution for large estates.
- Wave order: pilot (low risk, representative), then dependency-aligned waves; keep a shared database and its app tiers in the same wave.
- Agree RTO/RPO and a business case (TCO, licence portability, exit cost) before selecting tooling.
Worked example
A 400-VM estate has 120 web/app VMs, 30 Oracle Databases, and 250 batch/file servers. Dependency mapping shows the app VMs talk to three databases. Plan: wave 0 pilot = 8 web VMs on one non-critical app; wave 1 = those three databases with their app tiers (logical online replication to cap downtime); wave 2 = remaining apps; wave 3 = batch. Discovery flags 40 servers idling below 2% CPU - retire them, cutting scope ~10% and funding the pilot. Pattern: rehost the VMs, replatform the three databases to Exadata Database Service.
Traps that cost marks
- Sizing from allocated on-prem vCPUs instead of utilization, then paying for idle OCPUs.
- Skipping dependency mapping and migrating an app tier before its database.
- Treating "retire" as optional - assessments routinely find 10-20% idle assets.
- No agreed RTO/RPO, leaving the cutover plan with no defensible downtime budget.