Prepare and test your skills

Prepare and test your skills

Modernization-focused migration strategies change a workload while it moves so the workload can use cloud-native capabilities. A workload is an application, service, database, virtual machine, or container that runs in an environment and provides business value. Legacy workloads were built without the cloud in mind, so they can be hard to change, expensive to run, and unable to scale. Cloud-optimized workloads are natively scalable, portable, available, and secure. The goals of modernization are greater agility, lower infrastructure and operations costs, and higher reliability and resiliency.
Refactor, also called move and improve, modifies a workload to take advantage of cloud capabilities instead of just making it run in the new environment. Refactoring can happen during migration or before migration, and its goals are better performance, features, cost, and user experience. Some refactoring is required when the current architecture or infrastructure is not supported in the target environment, or when a major update is needed in addition to the move. Refactoring lets an application use platform features such as scalability, high availability, and portability. Refactor migrations take longer than rehost migrations and require learning new skills. Consider refactor when dependencies must be untangled, when the operating system, hardware, or database is unsupported, when compute or storage use is inefficient, or when deployment cannot be automated without extra effort.
Rearchitect, also called continue to modernize, changes how the code functions to optimize the workload for the cloud. Rearchitecting is similar to refactoring, but it restructures behavior and not just structure, which improves scalability, security, and agility. For example, a rearchitect migration can break one large monolithic workload into several independent microservices that run on Google Kubernetes Engine (GKE) (opens in a new tab) or Cloud Run. The microservices model lets each part scale and update independently. Rearchitecting is more complex than refactoring, so it takes more time and effort and can introduce bugs or security issues. Because of that risk, rearchitect migrations need several rounds of testing to confirm everything runs at its best. Rearchitect is one concrete way to reimagine an application around cloud-native design.
Rebuild, also called remove and replace, decommissions an existing application and completely redesigns and rewrites it as a fully cloud-optimized application. Choose rebuild when the current application no longer meets goals, is too costly to migrate another way, is not supported on Google Cloud (opens in a new tab), or is not worth maintaining. Rebuild is not suitable for ready-made applications because it requires rewriting the app from scratch. Rebuild is another concrete way to reimagine an application. A rebuild lets the new application use horizontal scalability, highly managed services, and high availability while removing the technical debt of the legacy version. The tradeoff is longer time than rehost or refactor, plus the need for new skills and new toolchains to provision, configure, and deploy the new environment.
Repurchase moves from a purchased on-premises workload to a cloud-hosted software-as-a-service (SaaS) equivalent. For example, an organization can move from on-premises collaboration software and local storage to Google Workspace. From a resources perspective, repurchase can be easier than refactoring, rebuilding, or rearchitecting because the organization does not rebuild the system itself. The tradeoffs are possible higher cost and less granular control over the cloud environment.
Each modernization strategy changes a different amount of the workload and offers different benefits for different effort:
| Strategy | What changes | Cloud benefits | Effort and time |
|---|---|---|---|
| Refactor | Modifies workload to use cloud capabilities | Performance, cost, user experience, scalability, high availability, portability | Longer than rehost; requires new skills |
| Rearchitect | Changes how code functions, such as monolith to microservices | Scalability, security, agility | More complex than refactor; needs extensive testing |
| Rebuild | Decommissions and rewrites app as cloud-optimized | Horizontal scalability, managed services, high availability; removes technical debt | Longer than rehost or refactor; needs new skills and toolchains |
| Repurchase | Moves to SaaS equivalent | Easier from resources perspective | May cost more; less granular control |
A phased approach often starts with lift and shift or replatform, then moves to refactor or rearchitect once the workload is stable in the cloud. Lifting and shifting first optimizes from an infrastructure perspective, and running in the cloud makes it easier to add cloud services later. Different workloads, and even different parts of one stack, can use different approaches. For example, a backend MySQL database can be replatformed to managed Cloud SQL (opens in a new tab), frontend virtual machines (VMs) can be refactored to containers on GKE Autopilot where Google manages nodes, scaling, and security, and hardware load balancing can be replaced with Cloud Load Balancing (opens in a new tab) and Google Cloud Armor.
Exam tip: Refactor (move and improve) modifies a workload to use cloud capabilities, rearchitect changes how the code functions, and rebuild replaces the application with a fully cloud-optimized rewrite.
Core migration pathways are the standard ways to move a workload to Google Cloud, from moving it unchanged to replacing parts of it with managed services. The two starting pathways are rehost, also called lift and shift, and replatform, also called lift and optimize. A common plan is to start with one of these pathways and continue to modernize later with refactor or rearchitect work. The three goals tied to this work are greater agility, lower infrastructure and operations costs, and higher reliability and resiliency.
Rehost, or lift and shift, moves a workload to Google Cloud largely as-is without changing its code or architecture. A typical rehost moves on-premises Windows Server, SQL Server, or VMware VMs to VM instances in Compute Engine (opens in a new tab). The main purpose is to shift the burden of physical servers to cloud infrastructure and make the workload cloud-ready for later optimization. Google Cloud tools for this pathway include Migrate to Virtual Machines, which moves VMs from on-premises and other clouds to Compute Engine, and VMware HCX, which moves VMs to Google Cloud VMware Engine. The operational impact is speed and low disruption because the application runs much as before. The team still manages the operating system and application but no longer manages physical hardware, power, cooling, and data-center facilities. Choose rehost when the workload meets one or more of these conditions:
Replatform, or lift and optimize, moves a workload to the cloud while replacing selected components with managed cloud services. The core application stays largely the same, but parts such as databases, runtime platforms, or load balancers are changed to work better in Google Cloud. For example, a self-hosted MySQL database can be replatformed to managed Cloud SQL (opens in a new tab), and VM-based workloads can be converted to containers on Google Kubernetes Engine (GKE) or Cloud Run. Tools for this pathway include Migrate to Containers, Datastream, and Database Migration Service, which help move to destinations such as Cloud SQL, Cloud Storage, and BigQuery. The operational impact is larger than rehost because configuration and integration work is required, but ongoing operations become simpler. Google manages tasks such as cluster configuration, scaling, security settings, patching, and replication for the replaced component. Choose replatform when straight lift and shift will not work well, when the architecture is not supported as-is, or as a first optimization step before breaking a monolith into microservices later.
Combining pathways means choosing the best approach for each workload and even for each part of one application stack. Business needs, technical requirements, and time and cost feasibility drive the choice for each part, so one approach does not have to fit everything. A phased timeline often starts with rehost or replatform and then moves to refactor or rearchitect once the workload is stable. Migrated applications can still communicate with other environments over a hybrid network connection.
| Pathway | What changes during the move | Google Cloud example |
|---|---|---|
| Rehost | Application and VM move unchanged; infrastructure shifts to cloud | On-premises VMware VMs move to Compute Engine with Migrate to Virtual Machines |
| Replatform | Application moves but selected parts are replaced with managed services | Self-hosted MySQL moves to Cloud SQL; VM workloads move to containers on GKE |
For example, one stack can split so the backend database is replatformed to Cloud SQL, the frontend VMs are refactored to containers on GKE Autopilot, and hardware load balancing and firewall functions are replaced with Cloud Load Balancing (opens in a new tab) and Google Cloud Armor.
Exam tip: Rehost moves VMs unchanged to Compute Engine, while replatform keeps the app but swaps a component for a managed service such as Cloud SQL.
Foundational migration terms describe what happens to each workload in a migration plan. A workload is the unit of migration, and each workload is assessed, categorized, and assigned a strategy such as retire, retain, or one of the move-and-modernize approaches. Retire, retain, and migrate decisions shape scope, waves, and connectivity needs.
A workload is any application, service, database, virtual machine, or container that runs in an environment and provides business value. Workloads can be legacy or cloud-optimized. Legacy workloads were developed without cloud environments in mind and are difficult to modify, expensive to run, and usually unable to scale. Cloud-optimized workloads are natively scalable, portable, available, and secure. When assessing a workload, teams evaluate characteristics that affect planning:
Dependencies and state matter most for difficulty. A workload with no dependencies is easier to migrate than one with many, a stateless workload rarely needs data migration while a stateful workload needs a data-move strategy, and workloads that tolerate downtime are simpler first movers than mission-critical workloads that need zero or near-zero downtime.
Retire means decommissioning a workload instead of migrating it, which removes it from migration scope. Retain means keeping a workload in its current environment for now rather than moving it to Google Cloud (opens in a new tab). Retain keeps the workload out of current migration waves but may require a hybrid network connection if workloads in Google Cloud still need to communicate with it. Choose retire when a workload no longer satisfies requirements, is not worth migrating, or can be replaced. Choose retain when technical, organizational, compliance, or data privacy restrictions prevent migration, or when new workloads go to the public cloud while existing workloads stay where they are.
| Decision | What happens to the workload | Effect on the migration plan |
|---|---|---|
| Retire | Workload is decommissioned and not moved | Removed from inventory, waves, and move groups; reduces scope and cost |
| Retain | Workload stays in its current environment | Excluded from waves; may need hybrid connectivity to migrated workloads |
| Migrate | Workload moves to Google Cloud using a strategy | Cataloged, grouped into waves and move groups, and migrated in sprints |
Migration strategies describe how a workload that will move gets to Google Cloud. Rehost (lift and shift) moves it as-is, replatform (lift and optimize) makes limited changes to use managed services, and refactor (move and improve) modifies it to use cloud capabilities. Re-architect (continue to modernize) changes how code functions for scalability, security, and agility, rebuild (remove and replace) decommissions and rewrites the app as cloud-optimized, and repurchase moves to a SaaS equivalent. Reimagine is the broad goal of redesigning around cloud-native capabilities, pursued through re-architect and rebuild.
| Strategy | What changes | When to choose it |
|---|---|---|
| Rehost | Workload moves as-is | Few dependencies, not worth refactoring, or third-party software |
| Replatform | Limited changes to use managed services | Architecture not supported as-is; for example, self-hosted MySQL to Cloud SQL |
| Refactor | Code or configuration modified for cloud capabilities | Dependencies must be untangled; inefficient compute or storage; deployment not automated |
| Re-architect | Code function changes, such as monolith to microservices on Google Kubernetes Engine (GKE) (opens in a new tab) or Cloud Run | Needs scalability, security, and agility |
| Rebuild | App decommissioned and rewritten | App no longer meets goals, too costly to migrate, or not supported on Google Cloud |
Repurchase moves from purchased on-premises software to SaaS, such as moving to Google Workspace (opens in a new tab). Repurchase can be easier on resources than refactoring or rebuilding, but it may cost more and offers less granular control.
A Google Cloud migration plan moves through four phases in order. First, Assess builds an inventory and catalogs workloads by business value, dependencies, teams, refactoring effort, licensing, availability, and downtime tolerance. Then, Plan groups workloads into waves, move groups, and sprints. Next, Deploy migrates the move groups in sprints. Finally, Optimize improves the migrated workloads with cloud services:
A wave is a group of applications that share characteristics or interdependencies, a move group is infrastructure resources and workloads that must migrate together, and sprints are the execution periods. Retired workloads are removed before grouping, retained workloads are excluded from waves, and good first movers are not business critical, have motivated teams, have few dependencies, are stateless, need minimal refactoring, and tolerate downtime. Research and development environments typically migrate first, followed by pre-production and production.
Exam tip: Retire removes a workload from migration by decommissioning it; retain keeps it in its current environment; rehost moves it to Google Cloud as-is.
Retire means decommissioning a workload instead of migrating it, which removes it from migration scope. Retain means keeping a workload in its current environment for now rather than moving it to Google Cloud, which excludes it from current migration waves but may require a hybrid network connection to migrated workloads.
Rehost, or lift and shift, moves a workload to Google Cloud largely as-is without changing its code or architecture, such as moving virtual machines to Compute Engine. Replatform, or lift and optimize, moves the workload while replacing selected components with managed cloud services, such as moving a self-hosted MySQL database to managed Cloud SQL, while the core application stays largely the same.
Refactor, also called move and improve, modifies a workload to take advantage of cloud capabilities such as scalability, high availability, and portability. Rearchitect, also called continue to modernize, changes how the code functions to optimize for the cloud, such as breaking a monolith into microservices, while rebuild, also called remove and replace, decommissions the existing application and completely rewrites it as a fully cloud-optimized application.
An organization is assessing its on-premises inventory to build a Google Cloud migration plan and identifies two specific workloads:
Which strategic migration decisions should the organization make for these workloads?
Last updated