Migration and data transfer

Migration moves applications, databases, and supporting resources from one environment to another, such as from an on-premises data center to a cloud platform. Data transfer is the movement of bytes that supports migration and ongoing operation. They are related but not identical: copying a database does not migrate application dependencies, permissions, monitoring, or operational responsibilities.

Cloud adoption starts with business goals and readiness assessment rather than copying servers immediately. Teams inventory workloads, map dependencies, identify regulatory requirements, and establish identity, networking, security, billing, and operational foundations. AWS describes adoption through the phases assess, mobilize, and migrate and modernize. Its Cloud Adoption Framework considers business, people, governance, platform, security, and operations. A practical approach uses a pilot followed by migration waves, grouping dependent applications to reduce disruption.

Migration strategies are commonly described as the seven Rs. Rehosting moves an application with minimal changes, while replatforming introduces targeted changes, such as moving a self-managed database to a managed database service. Refactoring or rearchitecting changes the application design more substantially. Repurchasing replaces an existing product, often with software as a service. Relocating moves an environment at the infrastructure or virtualization layer with minimal application changes. Retaining leaves a workload in its current environment, and retiring removes one that is no longer needed. The right choice depends on deadlines, licensing, compatibility, risk, and long-term operating costs; extensive modernization is not required for every migration.

Database migration requires both moving records and preserving application behavior. A homogeneous migration keeps the database engine type, while a heterogeneous migration changes engines and may require schema conversion, rewritten queries, and changes to stored procedures or data types. For a small database with an acceptable maintenance window, backup and restore or export and import may be sufficient. For a busy production system, an initial full load followed by change data capture can replicate ongoing changes while the source continues serving requests.

Replication reduces downtime but does not automatically eliminate it. For example, an order-processing application might replicate its source database to a managed target, validate records, briefly stop writes, wait for replication lag to clear, and then redirect connections. Teams must test compatibility, reconcile data, establish a rollback plan, and decide how to handle writes made after cutover. Replication also does not replace backups because unwanted changes can propagate to the target.

Transfer method depends on data volume, effective bandwidth, available time, security, and source capabilities. Online transfers can use internet connectivity or dedicated network connections. Offline transport can be appropriate when a supported physical-transfer option is available and network transfer would take too long. Estimate duration using actual achievable throughput, not just the advertised link speed, and allow for encryption, retries, small-file overhead, and continuing source changes.

Transfer pricing is directional and path-dependent. Incoming data transfer from the internet into AWS is generally free of data-transfer charges, but storage, requests, migration tools, and network processing can still cost money. Outgoing transfer to the internet is generally charged, subject to applicable allowances and service-specific exceptions. Cross-Region transfers generally incur charges, and replication can create recurring transfer costs as well as destination storage costs.

Within one Region, traffic is not automatically free. Same-Availability-Zone traffic using private addresses is often free for supported paths, whereas cross-Availability-Zone traffic commonly incurs charges, potentially on both sides. Service-specific exceptions apply, so identify the source, destination, Availability Zones, Regions, and network components before estimating costs. NAT gateways, private endpoints, and dedicated connections can add processing or connection charges independently of data-transfer pricing. A sound estimate includes the initial copy, ongoing synchronization, testing, backups, and normal traffic after migration.