The GCP CDL infrastructure modernization domain covers roughly 17% of the exam. The key questions are: which compute option fits a given workload, and how should existing systems be migrated to the cloud?
Migration Strategies — The 6 Rs
When an enterprise moves workloads to the cloud, it must decide the migration approach. Google Cloud defines six strategies:
| Strategy | Alias | Description | Example | |----------|-------|-------------|---------| | Rehost | Lift-and-shift | Move as-is, no changes | On-prem VM → Compute Engine VM | | Replatform | Lift, tinker, shift | Minimal changes, swap one component | Keep app code, replace DB with Cloud SQL | | Refactor | Re-architect | Redesign as cloud-native | Monolith → microservices + Cloud Run | | Retire | Decommission | Shut down unneeded systems | Remove zero-usage legacy servers | | Retain | Keep | Stay on-prem for now | Regulatory requirement | | Reimagine | Rebuild | Fully new cloud-native product | On-prem software → SaaS model |
Exam tip: "move without changes" = Rehost; "redesign for cloud" = Refactor.
!The 6 Rs of cloud migration strategy
Compute Options Compared
Google Cloud offers compute services at different levels of abstraction — from full control (IaaS) to minimal management (serverless).
| Service | Type | Control | Best For | |---------|------|---------|----------| | Compute Engine | VM / IaaS | High | OS-level control, legacy app migration | | GKE | Managed Kubernetes | Medium | Container orchestration, microservices | | Cloud Run | Serverless containers | Low | Stateless HTTP containers, scale-to-zero | | App Engine | Serverless PaaS | Very low | Web/mobile backends, auto-scaling | | Cloud Functions | Serverless functions | Minimal | Event-driven single-purpose tasks |
Compute Engine
Virtual machines running in Google's data centers. You manage the OS, software, and networking. Best when you need OS-level or kernel-level control, or are doing a lift-and-shift of a legacy application.
Spot / Preemptible VMs: Short-lived VMs that Google can reclaim at any time. In exchange, you get up to 91% cost savings. Ideal for batch jobs, data processing, or any fault-tolerant workload. Remember: "cheap but interruptible."
GKE — Google Kubernetes Engine
Kubernetes is the industry standard for managing containerized applications at scale. GKE is Google's managed Kubernetes service — it handles the control plane so you focus on your workloads.
GKE Autopilot: Fully managed mode where Google also manages the worker nodes. In Standard mode you configure and pay for nodes; in Autopilot you only define Pods and pay per Pod resource usage. Choose Autopilot when you want Kubernetes without infrastructure management overhead.
Cloud Run
Runs container images without managing servers. Scales to zero when idle (no cost), and scales up instantly on incoming requests. Billing is per request and CPU/memory used during processing.
Perfect for APIs or services with variable or unpredictable traffic — no idle costs. "Containers without servers."
App Engine
Fully managed platform for web applications and mobile backends. Upload code, Google handles infrastructure, scaling, and patching. Two environments: Standard (restricted runtimes, fastest scaling) and Flexible (custom runtimes via Docker).
Cloud Functions
Execute a single function in response to an event: file upload to Cloud Storage, Pub/Sub message, HTTP request, etc. Simplest serverless option — ideal for short, independent tasks.
Example: "When a file is uploaded to Cloud Storage, auto-generate a thumbnail" → Cloud Functions.
API Management
APIs are how services communicate. Google Cloud provides two API management services:
| Service | Target | Key Features | |---------|--------|--------------| | Apigee | Enterprise, external partner APIs | Gateway, analytics, security policies, developer portal, monetization | | Cloud Endpoints | Internal or small-scale APIs | Authentication, monitoring, gRPC/REST support |
Apigee is the full enterprise API platform — publish APIs to external developers, track usage, apply security policies, and monetize. Think of it as managing APIs as a product.
Cloud Endpoints is lighter — used mainly for APIs that run alongside other Google Cloud services.
Hybrid and Multi-Cloud
Many enterprises cannot fully migrate to cloud immediately. GKE Enterprise (formerly Anthos) enables unified management of Kubernetes-based applications across on-premises, AWS, Azure, and Google Cloud.
Key benefits: Avoid vendor lock-in: Stay flexible across cloud providers Leverage existing investments: Reuse on-prem infrastructure Single management console: One interface for all environments Consistent policy enforcement: Apply the same security and network policies everywhere
Exam Key Points
"Move as-is without changes" -- Rehost (lift-and-shift)
"Redesign as cloud-native" -- Refactor
"Minimal changes during move" -- Replatform
"VM with OS-level control" -- Compute Engine
"Container orchestration at scale" -- GKE
"Kubernetes without node management" -- GKE Autopilot
"Serverless containers, scale to zero" -- Cloud Run
"Event-driven single function" -- Cloud Functions
"Web app PaaS, upload code only" -- App Engine
"Cheap interruptible VM for batch jobs" -- Spot / Preemptible VM
"Enterprise API management, developer portal" -- Apigee
"Internal/simple API management" -- Cloud Endpoints
"On-prem + cloud unified management" -- GKE Enterprise (Anthos)