Start with the resource hierarchy, not the workloads
GCP structures resources as an organisation containing folders, which contain projects, which contain the actual resources. Getting this hierarchy right before migrating anything matters more on GCP than on some other providers, because IAM policies, budgets and organisation policies inherit down through folders — a well-designed hierarchy makes governance nearly automatic, and a poorly designed one is expensive to restructure later.
- Folders typically map to business units, environments (production, staging, development) or both, depending on how the organisation is structured.
- Projects are the actual billing and resource-isolation boundary — most teams end up with far more projects than they initially expect, and that is normal on GCP.
- Organisation Policy Service sets hard constraints (allowed regions, restricted APIs) that apply across the whole hierarchy from folders down.
Discovery: Migration Center
Google Cloud Migration Center provides a central place to assess an on-premise estate, import discovery data (including from third-party tools such as RVTools for VMware estates), and get sizing recommendations for target GCP resources based on that data — playing a broadly similar role to AWS Application Discovery Service.
Compute migration: Migrate to Virtual Machines
Migrate to Virtual Machines (formerly Velostrata) replicates on-premise or other-cloud virtual machines to GCP with continuous block-level replication, allowing a low-downtime cutover similar in concept to AWS MGN. It supports both a straightforward lift-and-shift and, in some cases, converting VMs to containers as part of the same migration wave for teams ready to modernise while they move.
Database migration: Database Migration Service and Datastream
Database Migration Service handles migrations into Cloud SQL and AlloyDB with minimal downtime for common source engines (MySQL, PostgreSQL, SQL Server). For streaming change data into BigQuery or other targets in near real time — useful for keeping an analytics environment current during a phased migration — Datastream provides change-data-capture from operational databases without needing to build custom replication pipelines.
Data platform migration: BigQuery Data Transfer Service and Dataflow
Where the migration goal includes moving an existing data warehouse, GCP has particularly strong tooling: BigQuery Data Transfer Service automates recurring data loads from other Google products, Amazon S3, Teradata and other warehouses directly into BigQuery, while Dataflow (built on Apache Beam) handles more complex batch and streaming transformation pipelines that go beyond a straightforward table copy. For teams whose migration is partly a data warehouse consolidation, this tooling is a genuine differentiator versus building custom extraction pipelines by hand.
Connectivity: Cloud VPN and Cloud Interconnect
Cloud VPN provides encrypted connectivity over the public internet and is adequate for smaller migrations or lower-throughput needs; Cloud Interconnect (Dedicated or Partner) provides a private, higher-throughput connection comparable to AWS Direct Connect or Azure ExpressRoute, and is worth planning for early given typical multi-week provisioning lead times.
Cost visibility: Recommender and committed use discounts
Active Assist Recommender surfaces machine-type right-sizing suggestions and idle-resource recommendations based on actual usage, the GCP equivalent of AWS Compute Optimizer. Committed use discounts function similarly to AWS Reserved Instances or Azure Reservations — apply them only once usage has stabilised after migration, not before, for the same reasons that apply on every provider.
Frequently asked questions
The top-level structure (key folders, environment separation) is worth settling before the first workload lands, since restructuring later is disruptive. Individual project creation can and should continue throughout the migration as new workloads are onboarded.
It performs a lift and shift by default, replicating VMs largely unchanged, but it also supports an option to migrate directly to containers for teams who want to modernise compute as part of the same project rather than as a separate phase later.
Continuous replication means the target database stays current with the source until the actual cutover moment, so downtime is limited to a short final switchover rather than the length of a full export and import.
BigQuery Data Transfer Service is purpose-built for recurring, scheduled loads from known source systems into BigQuery specifically. Dataflow is a general-purpose processing engine for more complex transformation logic that goes beyond a scheduled copy. Most data platform migrations use both, for different parts of the pipeline.
Provisioning timelines for a dedicated private connection (Cloud Interconnect, Direct Connect, ExpressRoute) are broadly comparable across all three providers and should be planned for weeks in advance regardless of which one is chosen.
No. As on every provider, commit only once usage has settled after right-sizing, typically a few weeks post-cutover. Committing early locks in whatever sizing mistakes exist at that point for the length of the term.
How we can help
Migration delivery across AWS, Azure or GCP, with a rollback plan for every step.
Data Architecture & EngineeringWarehouse and lakehouse design once the platform migration is complete.