Prepare and test your skills

Prepare and test your skills

Cloud transformation is the move from owning on-premises servers to running applications on Google Cloud (opens in a new tab) as a public cloud. Organizations make this move for business reasons such as lower cost and faster releases and for technical reasons such as global scale and automation. Together these drivers move companies from rigid hardware to elastic environments that are paid for as they are used.
Business and technical drivers for moving to the cloud are the outcomes and capabilities that justify migration. Business drivers include lower total cost of ownership, improved availability and resilience, faster development and testing, and innovation such as training machine learning models on existing data. Technical drivers include building analytics and artificial intelligence (AI), improving performance with multi-regional architectures, automating rollouts, using high-level application programming interfaces (APIs) and serverless services, and provisioning compute and storage in minutes instead of weeks.
Temporary needs also drive adoption, such as a three to six month project that needs large capacity, a multi-year transition that needs a hybrid and multicloud strategy (opens in a new tab), or a merger where two IT systems must be integrated quickly.
The shift from legacy infrastructure to cloud is the shift from fixed servers to agile architectures and from capital expenditure (CapEx) to operational expenditure (OpEx). A typical legacy system is a monolithic application built on .NET Framework, hosted on on-premises Windows Server, and connected to on-premises SQL Server. That design brings Windows licensing costs, difficult upgrades to a large codebase, and fixed capacity that is hard to budget and scale.
| Dimension | Legacy application | Modern cloud application |
|---|---|---|
| Management | Team manages servers, Windows, and SQL Server | Minimal overhead with managed services |
| Mobility | Tied to Windows with licensing costs | Portable with containers and no Windows dependency |
| Updates | Hard to upgrade a monolith | Fast upgrades with independent microservices |
| Scale and budget | Fixed hardware with little agility | Elastic serverless scaling with pay-as-you-go budgeting |
On-premises ownership means the customer manages cooling, physical security, hardware, virtualization, and applications. In public cloud the provider manages physical facilities, hardware, and the virtualization platform, so the customer focuses on applications and can use fully managed services. This makes short-term research and large distributed jobs practical because teams can create capacity quickly and shut it down when finished.
Exam tip: CapEx means buying servers you own with large upfront spending, while OpEx means paying on demand for what you use; the cloud driver is lowest possible CapEx with fast provisioning.
Migration is moving a workload to Google Cloud, while modernization is changing the application to use cloud-optimized capabilities such as managed services, containers, microservices, and serverless computing. Migration strategies describe how much changes during the move. Choosing a strategy depends on dependencies, cost, skills, and whether the technology can run in the cloud.
Lift-and-shift focuses on speed with minimal change, while re-architecting focuses on cloud-native design. Re-architecting takes more time, needs more testing, and requires new skills, but it unlocks independent scaling and faster releases. A common phased cycle is to first rehost or replatform to get into the cloud, then refactor or rearchitect once the workload runs stably and can connect over a hybrid network to other environments.
Different paths for different applications means each workload follows the migration approach that fits its requirements. Even parts of the same stack can use different approaches and still communicate. For example, a backend MySQL database can be replatformed to Cloud SQL while frontend VMs are refactored to containers on GKE Autopilot, where Google manages nodes, scaling, and security. Hardware load balancing and web application firewall functions can be replaced with Cloud Load Balancing and Google Cloud Armor.
Modernization also replaces self-managed parts with managed services that reduce operations work. Examples include Cloud Storage instead of network attached storage (NAS), Pub/Sub instead of self-hosted MSMQ, RabbitMQ, or Kafka, Identity-Aware Proxy (IAP) instead of custom-coded login, and Cloud Key Management Service (Cloud KMS) and Secret Manager instead of local secrets. A monolithic VM application can become independent microservices on Google Kubernetes Engine (GKE) or Cloud Run, and an API platform like Apigee can add security and analytics to legacy services with minimal code change.
Infrastructure modernization benefits are the scalability, reliability, and cost gains from moving from self-managed servers to managed compute, storage, and networking on Google Cloud (opens in a new tab). On-premises teams pay for power, cooling, and hardware whether capacity is used or not, while cloud teams pay based on consumption and scale up or down with demand. This shift moves patching, backup, and hardware repair to Google so teams focus on applications.
What improves when infrastructure moves to managed services is management overhead, portability, release speed, and budgeting. A legacy .NET monolith on Windows Server and SQL Server has licensing costs, limited deployment options, hard upgrades, and fixed capacity. A modernized architecture uses managed services for minimal overhead, .NET and containers without Windows dependency, microservices for independent upgrades, and serverless computing for elastic scaling.
The modernization path moves in three phases in order. First, rehost moves Windows Server and SQL Server as-is to Compute Engine VMs, manually or with tools such as Migrate to Virtual Machines. Second, replatform optimizes selected components such as database, caching, or storage without changing architecture and with minimal code changes. Third, re-architect and rebuild redesigns around Kubernetes, Linux containers, and serverless services. Different parts of the same system can be in different phases and still work together over a secure hybrid connection.
Faster releases, stronger security, and lower waste are the main business and operational outcomes of modernization. Modernized applications use independently deployable microservices, so one service can be updated without redeploying the whole monolith. Continuous integration tools such as Cloud Build and Cloud Deploy automate builds, run tests, provision environments, and scan for vulnerabilities within minutes.
Security and reliability improve because identity, secrets, and encryption move from application code to managed services. Application requests flow to Identity-Aware Proxy (IAP) for authentication instead of a custom login layer, while secrets flow from local files to Secret Manager and Cloud KMS. Traffic flows from users to Cloud Load Balancing, which distributes load across regions, while autoscaling follows demand and Cloud CDN caches content near users. Application logs flow to Cloud Logging and metrics flow to Cloud Monitoring, where dashboards combine server and application data for troubleshooting and planning.
Scaling automatically and paying only for what is used means capacity follows traffic instead of sitting idle. Right-sizing adjusts allocation to actual demand because over-provisioning wastes money and under-provisioning slows performance. Cloud Monitoring shows real-time use, Recommender suggests better allocations, and autoscaling adds or removes capacity without manual work.
Pricing options lower the rate paid for used capacity. Enterprise discounts apply to minimum total spend, resource-based committed use discounts (CUDs) apply to Compute Engine for one or three years, spend-based CUDs apply to product spend, and Spot VMs use unused capacity for fault-tolerant workloads at more than 80% lower cost than regular VMs. Network billing for ingress, egress, and cross-zone or cross-region traffic also affects cost, so placement across zones and regions affects high-availability design.
Exam tip: Automated scaling reduces what you use, while commitment discounts such as one-year or three-year CUDs reduce what you pay; choose scaling for efficiency and commitments for predictable use.
Application runtimes and containerization choices determine how much control the team keeps and how applications scale. Containerization packages an application and dependencies into a lightweight container that runs consistently across environments. Containers are lighter than VMs because they do not include a full operating system, so multiple containers can pack onto the same host for greater density. Migrate to Containers can extract container images from VM workloads for GKE or Cloud Run without source code or a full rewrite.
| Feature | GKE | Cloud Run |
|---|---|---|
| Management | Customer manages Kubernetes clusters and nodes while Google manages infrastructure | Customer provides code or image and Cloud Run runs it with no cluster to manage |
| Control | Full control over nodes, networking, policies, and add-ons | Simple settings for concurrency, variables, and connections |
| Application types | Stateless and stateful and complex microservices | Stateless request-driven and event-driven web services |
| Pricing | Pay per cluster per hour | Pay per use rounded to nearest 100 milliseconds |
Replatforming keeps the architecture but swaps components with minimal configuration, such as Cloud SQL instead of self-hosted SQL Server, MySQL, or Postgres, Memorystore instead of self-hosted Redis, and Managed Service for Microsoft Active Directory instead of self-hosted Active Directory. Refactoring breaks a monolith into microservices on GKE or Cloud Run for independent development and deployment. Frontend components are often refactored first because frontends are often stateless, change frequently, and benefit most from load balancing, multiregional deployment, and serverless execution.
Foundational migration terms describe what happens to each workload in a migration plan. A workload is an application, service, database, virtual machine, or container that runs in an environment and provides business value. Each workload is assessed, assigned a decision such as retire, retain, or migrate, and grouped into waves for execution.
What a workload is and how retire, retain, and migrate shape scope determines which systems move and which do not. Legacy workloads were built without the cloud in mind and are hard to change, expensive to run, and unable to scale. Cloud-optimized workloads are natively scalable, portable, available, and secure. Assessment looks at business value, criticality, teams, dependencies, refactoring effort, compliance and licensing, availability needs, and allowable downtime.
| Decision | What happens to the workload | Effect on the plan |
|---|---|---|
| Retire | Workload is decommissioned and not moved | Removed from inventory, waves, and cost |
| Retain | Workload stays in its current environment | Excluded from waves; may need hybrid connectivity |
| Migrate | Workload moves to Google Cloud (opens in a new tab) with a strategy | Grouped into waves and move groups and migrated in sprints |
Choose retire when a workload no longer meets requirements or is not worth migrating. Choose retain when technical, compliance, or privacy limits prevent migration. A workload with no dependencies is easier to move 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 downtime.
Rehost and replatform as starting migration pathways are the ways to move with little change or with limited optimization. Rehost, or lift and shift, moves a workload largely as-is without changing code or architecture. A typical rehost moves on-premises Windows Server, SQL Server, or VMware VMs to Compute Engine VM instances. Tools include Migrate to Virtual Machines and VMware HCX for moves to Google Cloud VMware Engine. The team still manages the operating system and application but no longer manages physical hardware, power, and cooling.
Replatform, or lift and optimize, keeps the core application but replaces selected parts with managed services. For example, self-hosted MySQL moves to Cloud SQL and VM workloads move to containers on GKE or Cloud Run. Tools include Migrate to Containers, Datastream, and Database Migration Service for destinations such as Cloud SQL, Cloud Storage, and BigQuery. Google then manages patching, scaling, security settings, and replication for the replaced component. Choose rehost when dependencies are few, refactoring is not feasible, or third-party software limits change. Choose replatform when lift-and-shift will not run well as-is or as a first optimization before microservices.
Refactor, rearchitect, rebuild, and repurchase for modernization are strategies that change the workload to gain cloud capabilities. Refactor, also called move and improve, modifies a workload to use scalability, high availability, and portability for better performance, cost, features, and user experience. Consider refactor when dependencies must be untangled, the operating system or database is unsupported, compute or storage use is inefficient, or deployment cannot be automated. Refactor takes longer than rehost and requires new skills.
Rearchitect, also called continue to modernize, changes how code functions, such as breaking a monolith into independent microservices on GKE or Cloud Run. Rearchitect improves scalability, security, and agility but is more complex than refactor and needs extensive testing. Rebuild, also called remove and replace, decommissions the old application and rewrites it as fully cloud-optimized with horizontal scalability and managed services while removing technical debt. Rebuild takes longest and needs new toolchains. Repurchase moves to SaaS such as Google Workspace, which is easier on resources but can cost more with less granular control. Reimagine is the broad goal of redesigning around cloud-native capabilities through rearchitect and rebuild.
Exam tip: Retire removes a workload by decommissioning it, retain keeps it where it is, and rehost moves it to Google Cloud as-is; refactor modifies it, rearchitect changes how its code functions, and rebuild replaces it.
Organizing the move with Assess, Plan, Deploy, and Optimize is the ordered migration process that groups workloads for execution. Discovery and assessment stay ongoing while risk assessment runs in parallel. Starting with a workload that is not business-critical, has minimal dependencies, and can serve as a blueprint helps establish patterns for later waves.
A migration wave is a group of applications that share characteristics or interdependencies. Within a wave, a move group is infrastructure and workloads that must migrate together, and sprints are the execution periods. Research and development environments typically migrate first, followed by pre-production and production. A phased approach often starts with rehost or replatform, then moves to refactor or rearchitect once workloads are stable in the cloud.
Capital expenditure (CapEx) means buying servers you own with large upfront spending, while operational expenditure (OpEx) means paying on demand for what you use. Cloud transformation shifts organizations from owning on-premises hardware with upfront CapEx to using elastic cloud resources billed as OpEx for what is used.
Retire means decommissioning a workload that no longer meets requirements or is not worth migrating, retain means keeping a workload in its current environment when technical, compliance, or privacy limits prevent migration, and migrate means moving a workload to Google Cloud with a migration strategy. Retired workloads are removed from inventory, waves, and cost, retained workloads are excluded from waves and may need hybrid connectivity, and migrated workloads are grouped into waves and move groups and migrated in sprints.
Rehost (lift and shift) moves a workload largely as-is without changing code or architecture, such as to Compute Engine virtual machines, while replatform (lift and optimize) keeps the core application but replaces selected parts with managed services such as Cloud SQL. Refactor (move and improve) modifies a workload to use scalability, high availability, and portability, rearchitect changes how code functions such as breaking a monolith into independent microservices on Google Kubernetes Engine (GKE) or Cloud Run, rebuild decommissions and rewrites the application as fully cloud-optimized, and repurchase moves to software-as-a-service (SaaS) such as Google Workspace.
Last updated