This is Part 4 of a blog series about Google Cloud Professional Cloud Architect certification. Part 1 focuses on my experience — how I prepared and what the exam was like the first time. Part 2 covers my renewal experience with AI centred exam. Part 3 shares the resources I used and my notes on GCP products. Part 4 includes my notes on case studies.
These are my personal notes on publicly available Google Cloud PCA case studies. They do not contain real exam questions. In the sections below I use phrases like “possible services to evaluate,” “reasonable options include,” “one likely fit is,” and “trade-offs to consider” — as study aids to reason about scenarios, not definitive answer keys. Always read the actual case study PDFs and documentation to form your own view.
Altostrat
Company: Media industry: podcasts, interviews, news broadcasts, documentaries (content management system, media lib).
Solution: GenAI for personalised recommendations, natural language interactions, seamless self-service support. Dynamic pricing, targeted marketing, personalised product suggestions.
Tech env
- GKE to serve, Cloud Storage for media objects.
- BigQuery for analytics.
- For video transcoding, metadata extraction, personalised recommendations: possible services to evaluate include Cloud Run / Cloud Functions; Transcoder API (managed, scaled UHD ~$0.06/min); Cloud Batch for cost reduction; Cloud Run Job for transcoding. Trade-offs to consider: managed vs. self-managed, cost vs. flexibility.
- Legacy: on-premises for content ingestion and archival.
- User auth: Google identity and third-party providers.
- Monitoring: Google Monitoring and Prometheus, alerts via email.
Business and tech requirements — possible approaches
- Reliability across env (GCP + on-prem). Reasonable options include Interconnect for hybrid connectivity; Cloud Composer (Airflow) or Workflows for orchestration. Trade-offs to consider: managed vs. self-hosted orchestration.
- Simplified and rapid deploy, CI/CD with centralised management. Possible services to evaluate: Cloud Build (CI) with security scans and tests; Artifact Registry (images, vulnerability scanning); Cloud Deploy (CD). For hybrid and centralised management, one likely fit is Anthos/GKE Fleet.
- Optimise storage costs / high-perf hybrid connectivity for data ingestion. Options include Cloud Storage with storage classes, lifecycle rules, and archive; CDN for availability and scalability. For hybrid high throughput, consider Dedicated or Partner Interconnect.
- NL support 24/7, LLMs and AI for personalised experiences. Reasonable options include Vertex AI Agent Builder (chat, tools, grounding) and Dialogflow CX.
- NL summaries / automated summarization for diverse media. One possible flow: Speech-to-Text (Chirp3: sync, async, stream) → Vertex AI (Gemini) on transcript; Eventarc on new upload → Cloud Run → store summaries in BigQuery or Firestore and index. Trade-offs to consider: batch vs. event-driven, storage choice for summaries.
- Vision and NL for metadata. Possible services to evaluate: Video Intelligence API, Vision API, Natural Language API; Gemini for richer understanding.
- AI filters for content. Reasonable options include Vertex AI with safety settings and human-in-the-loop; Model Armor for additional protection.
- Media analysis for trends and insights. One likely fit is BigQuery with Looker and Dataflow for streaming aggregation.
- Analytics dashboards. Reasonable options include BigQuery and Looker.
Cymbal
Company: Retail with extensive product catalog.
Solution: GenAI for product catalog: attributes, description, images, improve consistency. Google Cloud's Discovery AI agents as personalised sellers. Improve security, costs, reliability, monitoring with cloud solutions.
Tech env
- Mix of on-prem and cloud, Kubernetes clusters. Reasonable options include Anthos/GKE Fleet and Interconnect for hybrid; VPN is a possibility if bandwidth requirements are lower. Trade-offs to consider: Interconnect vs. VPN (latency, cost, throughput).
- MySQL, Microsoft SQL Server: one likely fit is Cloud SQL; Redis → Memorystore; MongoDB: Firestore is a possible migration target if you have capacity to move workloads there.
- Legacy file-based integrations with on-prem. Possible services to evaluate: Cloud Storage for files; ETL batch — Pub/Sub → Workflows/Composer → Dataflow; SFTP-style transfers — signed URLs to GCS or Cloud Run with HTTPS; Storage Transfer Service for batch moves.
- IVR (Interactive Voice Response) to route customer calls. Reasonable options include CX Agent Studio, Dialogflow CX for voice and chat; CCAI patterns (agent assist, call summarization with Vertex AI); backend integration via Cloud Run; Agent Assist for human agents.
- Monitoring (e.g. Grafana, Nagios, Elastic). One likely fit is Cloud Operations Suite.
Business and tech requirements — possible approaches
- Automate product catalog enrichment, attributes, and image generation with human-in-the-loop (HITL) review. One possible flow: GCS landing → Pub/Sub trigger → Cloud Run enrichment worker → Vertex AI (text/image) → HITL approval UI → write to catalog DB; analytics with BigQuery. Trade-offs to consider: event-driven vs. batch, where to run HITL UI.
- Improve product discoverability, search relevance, and natural-language product results. Reasonable options include Discovery Engine or Matching Engine; Vertex AI Search for commerce; Dialogflow CX (or an agent layer) calling a search tool.
- Increase customer engagement: interactive, personalised, intuitive experience. One likely fit is Dialogflow CX plus search/recommendation fulfillment and Vertex AI Search for commerce.
- Reduce call center staffing and data-center hosting costs. Possible services to evaluate: Google Contact Center AI (CCAI), CX Agent Studio, Dialogflow CX for voice and chat.
Two years ago I used thecertsguy.com, but I think the website doesn’t work anymore. This is a summary I made for myself based on that website. Some maybe outdated, I will check them, hope soon.
EHR Healthcare
- Maintain legacy interfaces to insurance providers (on-prem + cloud). Reasonable options include VPN/Interconnect, BYOIP. Trade-offs to consider: bandwidth, SLA, colocation.
- Decrease infrastructure administration costs. One likely fit is managed services (e.g. Cloud SQL, GKE, Dataflow).
- Consistent management of container-based customer-facing apps. Possible services to evaluate: Anthos.
- Secure, high-performance connection on-prem ↔ GCP. One likely fit is Dedicated Interconnect.
- Consistent logging, retention, monitoring, alerting. Reasonable options include Anthos Service Mesh and Cloud Operations Suite.
- Maintain and manage multiple container-based environments. One likely fit is Anthos.
- Dynamically scale and provision new environments (IaC). Possible services to evaluate: Terraform.
- Ingest and process data from new providers. Reasonable options include Pub/Sub, Dataflow.
- Migrate non-containerized legacy apps. One likely fit is Migrate for Anthos.
- Single pane of glass for containers. Possible approaches: Anthos Service Mesh and Anthos Config Management.
- Reduced latency for content. One likely fit is Cloud CDN.
- MySQL, MS SQL Server: Cloud SQL; Redis: Memorystore; MongoDB: Firestore — reasonable migration targets to evaluate.
- Microsoft AD integration. Possible approaches: Google Cloud Directory Sync and AD FS.
- Monitoring & SRE. Reasonable options include Cloud Ops; Anthos Service Mesh for multi-platform containers; chaos engineering and Istio fault injection; “LETS” metrics (Latency, Errors, Traffic, Saturation) with percentiles to reduce alert fatigue.
Summary (services that often appear in this scenario): Anthos, Migrate for Anthos, Cloud CDN, Cloud SQL, Memorystore, Pub/Sub, Dataflow, BigQuery, Datastudio/Looker, AI Platform, ADFS/GCDS, Cloud Operations Suite, Anthos Service Mesh, Terraform, Cloud Build, Deploy Manager, Jenkins (CI), Spinnaker (CD).
Mountkirk Games
- Goals: multiple gaming platforms (mobile, desktop, tablets), multiple regions, rapid iteration (CI/CD), minimise latency, dynamic scaling, managed services, minimise costs. CDN is often part of the picture for latency.
- Scale based on game activity. One likely fit is Kubernetes. Near real-time global leaderboard: Memorystore. Game activity logs for analysis: Cloud Logging, GCS. GPU rendering server-side: Kubernetes. Migrate legacy games: trade-offs to consider between lift-and-shift and re-platform.
- Hundreds of users joining simultaneously globally. Reasonable options include Kubernetes backends in multiple regions and Global HTTPS LB to balance traffic.
- User profiles. One likely fit is Firestore; leaderboards → Cloud Memorystore.
- Game activity logs. Possible flow: Cloud Logging → GCS; analyse later with BigQuery.
- DevOps / rapid iterations. Reasonable options include Terraform and Cloud Build.
Summary (services that often appear): Kubernetes (game server), Firestore (user profiles), Memorystore (leaderboards), Cloud Logging (game activity logs), GCS (log storage), Terraform & Cloud Build (rapid iterations).
Helicopter Racing League
- Increase telemetry and insights. Reasonable options include IoT Core, BigQuery, Looker/Datastudio.
- Content serving from regions closer to viewers. One likely fit is GCS (raw + transcoded, multi-region) with CDN.
- Store unstructured video. GCS; after transcoding, GCS → AI Platform (or Vertex AI) for predictions is a possible flow.
- Increased telemetry, real-time analytics and predictions. Possible services to evaluate: Pub/Sub, Cloud Dataflow → AI Platform (or Vertex AI).
- Real-time video analysis and transcoding. One possible flow: GCS; Cloud Function on upload → Transcoder API → GCS; Video Intelligence API. Trade-offs to consider: Transcoder API vs. self-managed, when to use Video Intelligence.
- Expose prediction model to partners. Reasonable options include Apigee or Cloud Endpoints (Apigee if monetisation is in scope).
- Minimise operational complexity; maintain or increase prediction throughput and accuracy. One likely fit is BigQuery ML.
Summary (services that often appear): IoT Core, BigQuery, Looker/Datastudio, AI Platform, Transcoder API, GCS, Video Intelligence API.
TerramEarth
- Decrease cloud operational costs and adapt to seasonality. Reasonable options include IoT Core, Pub/Sub, Dataflow to decouple ingestion and processing. Trade-offs to consider: serverless vs. fixed capacity for seasonal spikes.
- Increase speed and reliability of dev workflow. Possible services to evaluate: Cloud Build, Deployment Manager, Cloud KMS, Secret Manager, Cloud Operations (e.g. audit and network logs, VPC Flow Logs).
- Remote developer productivity. One likely fit is Identity-Aware Proxy (IAP); sandbox project in a separate folder with IAM and network policies.
- Custom API services for dealers and partners. One likely fit is Apigee.
- Interconnect with private data center. Reasonable options include Cloud Router plus Interconnect (Partner or Dedicated).
Summary (services that often appear): AI Platform, IoT Core, Pub/Sub, Dataflow, Terraform, Deployment Manager, Cloud KMS, Secret Manager, Cloud Operations Suite, Cloud IAM, IAP, Apigee.