Google Cloud Professional Cloud Architect Designing and Planning a Cloud Solution Architecture Flashcards

24 sample cards from the Designing and Planning a Cloud Solution Architecture deck

Sampled from the end of the deck, so these are different cards from the ones shown on the syllabus page.

1. Compare Cloud VPN, Dedicated Interconnect, and Partner Interconnect for hybrid connectivity.

   Cloud VPN uses encrypted IPsec tunnels over the public internet (lower cost, lower bandwidth). Dedicated Interconnect is a direct physical connection to Google (10/100 Gbps, lowest latency, highest cost). Partner Interconnect connects via a service provider for bandwidths between VPN and Dedicated.

2. What is the role of Cloud Load Balancing tiers in topology design (global vs regional)?

   Global external load balancers (e.g., global external Application LB) distribute traffic across multiple regions with a single anycast IP for low-latency global access. Regional load balancers serve traffic within one region. Choice depends on whether the backend is multi-region or single-region.

3. What are quotas in Google Cloud and how do they differ from limits?

   Quotas are configurable caps on resource consumption per project/region (e.g., number of CPUs, API requests/min) that protect against overuse and runaway costs and can be raised via request. Limits are fixed system maximums that cannot be changed.

4. What is capacity planning and what key inputs does it require?

   Capacity planning estimates the resources needed to meet demand at acceptable performance/cost. Inputs include current/forecast traffic, peak vs average load, growth rate, per-request resource cost, latency targets, and headroom for spikes and failover.

5. Why must quota planning precede a large migration or launch?

   Because default quotas (CPUs, IP addresses, API rate) may be insufficient for the target scale; quota increase requests can take time to approve. Failing to plan causes failed deployments or throttling at cutover.

6. What is workload assessment/discovery in a migration, and what tools support it?

   It is the inventory and analysis of existing applications, dependencies, performance, and resource usage to plan migration. Tools include Migration Center (formerly StratoZone/migration assessments) for discovery, dependency mapping, right-sizing, and TCO estimation.

7. What are the '6 Rs' of migration strategy?

   Rehost (lift-and-shift), Replatform (lift-and-optimize), Refactor/Re-architect, Repurchase (move to SaaS), Retire (decommission), and Retain (keep on-prem for now).

8. Define rehost, replatform, and refactor and their relative effort/benefit.

   Rehost: move as-is to VMs (fast, low effort, least cloud-native benefit). Replatform: minor optimizations (e.g., managed DB) for moderate benefit. Refactor: re-architect to cloud-native/serverless/containers (highest effort, greatest long-term agility and cost benefit).

9. What Google Cloud tool migrates VMs from on-prem/other clouds, and what migration style does it enable?

   Migrate to Virtual Machines (formerly Migrate for Compute Engine) performs agentless VM migration with streaming/lift-and-shift, allowing test-clone before final cutover to minimize downtime.

10. Compare online (live) versus offline data migration approaches.

    Online migration continuously replicates data while the source stays live, enabling minimal-downtime cutover (e.g., Database Migration Service CDC). Offline migration copies a static snapshot during a maintenance window (simpler, but requires downtime and the source is read-only/quiesced).

11. When is Transfer Appliance preferred over online Storage Transfer Service?

    Transfer Appliance (physical shippable device) is preferred for very large datasets (hundreds of TB to PB) where network transfer would be too slow or costly. Online transfer suits smaller datasets or when bandwidth and time are sufficient.

12. What is Database Migration Service and what makes it low-downtime?

    DMS is a managed service that migrates databases (e.g., to Cloud SQL/AlloyDB) using continuous change data capture (CDC) replication, keeping the destination in sync with the live source until cutover, minimizing downtime and data loss.

13. What is a cutover plan and what must it include?

    A cutover plan defines the steps to switch production from source to target system, including timing/maintenance window, data sync verification, DNS/traffic redirection, freeze of source writes, validation checks, success criteria, and a defined rollback trigger.

14. Why is rollback planning essential in a migration, and what does it require?

    Rollback lets you revert to the source if the cutover fails validation, preventing extended outage or data loss. It requires keeping the source intact/recoverable, a tested reversion procedure, preserved data sync direction, and clear go/no-go criteria.

15. How does validation testing differ before versus after cutover?

    Pre-cutover validation tests the migrated environment in parallel (functional, performance, data integrity) without serving production. Post-cutover validation confirms the live system meets success criteria under real traffic before declaring success and decommissioning the source.

16. Compare Google Cloud pricing models: on-demand, committed use discounts, and spot/preemptible VMs.

    On-demand: pay-as-you-go, full price, max flexibility. Committed Use Discounts (CUDs): 1- or 3-year commitment for ~25-57% savings on steady workloads. Spot VMs: deeply discounted (up to ~60-91%) but can be preempted, suited to fault-tolerant batch jobs.

17. What is a Sustained Use Discount in Compute Engine?

    A Sustained Use Discount is an automatic discount applied to eligible VMs that run for a significant portion of the billing month (the more of the month they run, the larger the discount, up to ~30%), requiring no commitment.

18. What tools support cloud cost estimation and ongoing cost optimization in Google Cloud?

    The Google Cloud Pricing Calculator for upfront estimates; Billing reports, budgets/alerts, and Cost Table for tracking; and Recommender (rightsizing, idle resource, CUD recommendations) plus the FinOps practice for ongoing optimization.

19. What is right-sizing and how does the Recommender help?

    Right-sizing matches provisioned resources (machine type, disk, memory) to actual utilization to remove waste. Compute Engine's rightsizing Recommender analyzes historical metrics and suggests smaller/larger machine types to cut cost while meeting performance needs.

20. How does autoscaling support designing for elasticity?

    Autoscaling automatically adds/removes resources (e.g., Managed Instance Group autoscaler, Cloud Run/GKE autoscaling) based on metrics like CPU, load-balancer utilization, or custom signals, so capacity tracks demand—improving cost efficiency and handling spikes without manual intervention.

21. What architectural practices make a system able to evolve over time?

    Loose coupling (microservices, APIs, async messaging via Pub/Sub), infrastructure as code, statelessness, modular boundaries, and managed services. These reduce blast radius and let components be replaced/scaled independently as requirements change.

22. What is a continuous improvement roadmap in cloud advisory, and what does it typically include?

    It is a forward-looking plan to keep optimizing after initial deployment, covering iterative cost optimization (FinOps), reliability/SLO refinement, security posture improvements, modernization (containerize/serverless), and automation maturity, reviewed on a recurring cadence.

23. What is the role of a decision framework when advising on customer success?

    A decision framework provides structured, criteria-based evaluation (e.g., weighing cost, risk, time-to-value, compliance, and business impact) so trade-offs are transparent and aligned to the customer's priorities, improving buy-in and measurable success outcomes.

24. How do availability tiers map to architecture decisions (single-zone vs multi-zone vs multi-region)?

    Higher availability tiers require broader fault-domain isolation: single-zone for low/dev tiers, multi-zone (regional) for production high availability (~99.9-99.99%), and multi-region for the highest tiers and DR (>99.99%). Each step adds cost and complexity, so tier is driven by the SLO.

What this deck covers

Designing and Planning a Cloud Solution Architecture flashcards FAQ