Cloud migration cutovers
Budget for the period when both environments run
Migration costs peak during overlap. Include replication, validation and recovery capacity before using the target's steady-state bill to justify the window.
In this article
Separate temporary and ongoing costs
The target environment may be cheaper to operate after retirement of the source, while the migration itself temporarily costs more. Show those periods separately.
Include source and target compute, data copies, replication infrastructure, transfer and the engineering effort needed for reconciliation. Use current provider pricing for an actual estimate rather than borrowing a generic rate.
Also account for contractual commitments that remain payable after traffic moves. Shutting down a server does not necessarily end its associated commercial cost.
Estimate the transfer window
Measure effective transfer and change-application rates using representative data. Available network bandwidth alone does not determine completion time if the target cannot ingest changes as quickly.
Compare the rate of new source changes with the sustained catch-up rate. If incoming work equals or exceeds the target's effective application capacity, the gap may never close during ordinary traffic.
Use this evidence to decide whether to increase capacity, reduce source activity or allow a longer write pause. Each option changes the business interruption and cost.
Keep recovery resources available
Retaining the source for an observation period costs money, but retiring it immediately can remove an important recovery option. Define what capability must remain and for how long.
A powered-off source with incomplete access or outdated data may not provide the recovery value assumed in the budget. Tie the retained resources to a tested recovery procedure.
Likewise, target capacity should cover initial cache warming and resumed queues. A deployment sized only for an already warm steady state may struggle immediately after the switch.
Put an owner on retirement
Set the conditions for removing replication jobs, temporary storage and old infrastructure. Include business validation and recovery-window completion.
Track those conditions rather than leaving cleanup as an informal future task. Long-running overlap can erase expected savings while increasing operational complexity.
The useful comparison is the total cost of a controlled transition and stable operation. A smaller target bill is only one part of that calculation, and it should not force a cutover window that cannot accommodate verification or recovery.
Primary sources
AWS DMS: ongoing replicationReferences checked 11 September 2026.