The Google Cloud Professional Cloud Architect (PCA) exam is less about knowing which buttons to click and more about business-to-technical mapping. In 2026, the exam heavily features Vertex AI, GKE Autopilot, and multi-regional disaster recovery.
This cheat sheet covers the core pillars of the 2026 syllabus.
1. Architecture Design (The "Business" Pillar)
The PCA exam always starts with business requirements. You must choose the solution that is cheapest, fastest, or most reliable based on the prompt.
Migration Strategies
- Lift & Shift (Rehost): Move VMs as-is to Compute Engine using Migrate to Virtual Machines.
- Improve & Move (Replatform): Containerize apps and move to GKE or Cloud Run.
- Rip & Replace (Refactor): Rewrite for cloud-native services like Cloud Spanner or Cloud Functions.
Case Studies (20-30% of Exam)
Expect 2 case studies (e.g., EHR Healthcare, TerramEarth, Mountkirk Games).
- Key Tip: If the company is in Healthcare/Finance, prioritize Compliance (HIPAA/PCI) and VPC Service Controls.
- Key Tip: If it's a Gaming/IoT company, prioritize Global Scalability (Spanner) and Low Latency (Global Load Balancing).
2. Compute: Choosing the Right "Brain"
| Service | Best Use Case | Key Feature |
|---|---|---|
| Compute Engine | Legacy apps, custom OS, specific hardware. | Spot VMs for 60-91% cost savings. |
| GKE (Autopilot) | Microservices, Kubernetes-native. | Google manages nodes; you pay for Pods. |
| Cloud Run | Web APIs, serverless containers. | Scales to zero; handles request-based traffic. |
| Vertex AI | Generative AI, ML training/hosting. | Model Garden & Agent Builder (New for 2026). |
3. Storage & Databases
Choosing the wrong database is the most common reason for architectural failure.
- Cloud Storage: Object storage. Use Lifecycle Policies to move from Standard -> Nearline -> Coldline -> Archive.
- Cloud SQL: Managed MySQL/Postgres/SQL Server. Vertical scaling only.
- Cloud Spanner: Unlimited horizontal scale + Strong Consistency. Use for global finance apps.
- Bigtable: High-throughput NoSQL for IoT/Time-series. (Think: Millions of reads/writes per sec).
- BigQuery: Serverless Data Warehouse. Use for Petabyte-scale analytics.
4. Networking & Security
The "Glue" that keeps the architecture secure.
Load Balancing
- Global External HTTP(S) LB: Uses a single Anycast IP. Routes traffic to the closest region.
- Cloud Armor: WAF for protecting against DDoS and SQL injection.
- Identity-Aware Proxy (IAP): Access apps without a VPN (Zero Trust).
Hybrid Connectivity
- Cloud VPN: Encrypted, over the public internet (Fast to setup).
- Dedicated Interconnect: Physical connection (10G/100G). Highest reliability and lowest cost for huge data.
- VPC Service Controls: Create a "security perimeter" to prevent data exfiltration.
5. Generative AI & Data (2026 Focus)
The 2026 syllabus expects Architects to understand AI integration.
- Vertex AI Search & Conversation: Quickly build RAG (Retrieval-Augmented Generation) apps.
- BigQuery ML: Run ML models directly inside BigQuery using SQL.
- Pub/Sub: The messaging backbone for asynchronous, event-driven architectures.
6. Reliability & Operations
- RTO (Recovery Time Objective): How quickly must you be back up? (Lower = More expensive).
- RPO (Recovery Point Objective): How much data can you lose? (Zero = Multi-region Spanner/SQL).
- Cloud Logging/Monitoring: Create SLIs (Indicators) and SLOs (Objectives) to measure success.
📝 PCA "Pro-Tips" for 2026
- Cost Optimization: Always choose Committed Use Discounts (CUDs) for stable workloads.
-
Organization Policy: Use this to restrict which regions developers can deploy to (e.g.,
constraints/gcp.resourceLocations). - IAM: Always follow the Principle of Least Privilege. Use Service Accounts for app-to-app communication.
- Binary Authorization: Ensure only trusted container images are deployed to GKE.
Good luck with your PCA Certification!
Top comments (0)