Google Cloud Platform (GCP) Interview Guide: Cloud Engineer and Architect Questions
GCP interviews appear when you're applying to Google-adjacent companies, cloud engineering roles at enterprises that have standardized on Google Cloud, or Google itself. The questions split between GCP-specific services, cloud-agnostic architecture, and data engineering (GCP is strong in data). Here's how to prepare.
Why GCP? Know the Positioning
Interviewers often open with "why Google Cloud over AWS/Azure?" Have a genuine answer: GCP excels at data analytics (BigQuery), machine learning infrastructure (Vertex AI, TPUs), Kubernetes (Google invented it; GKE is the reference implementation), and global networking (private fiber backbone). Companies choose GCP when they're heavy on analytics or already in the Google ecosystem (Workspace, etc.).
Core Compute: GKE and Cloud Run
GKE (Google Kubernetes Engine): The managed Kubernetes service. Interview topics: autopilot vs. standard mode (autopilot manages node pools, standard gives more control), workload identity for pod-level IAM, vertical and horizontal pod autoscaling, multi-cluster setups with Anthos.
Expect the question: "How do you handle pod IAM permissions in GKE?" Answer: Workload Identity — bind a Kubernetes service account to a GCP service account, allowing pods to authenticate to GCP services without credentials in environment variables.
Cloud Run: Fully managed serverless container platform. Scales to zero, charges per request. Best for stateless HTTP services, event-driven workloads, and microservices with variable traffic. Compare with App Engine: Cloud Run is container-based and more flexible; App Engine is more opinionated. Cloud Run vs. GKE: Cloud Run for simplicity and auto-scaling to zero; GKE for stateful workloads, complex networking, and when you need more control.
Storage: GCS, Cloud SQL, Bigtable, Spanner
Cloud Storage (GCS): Object storage. Key interview topics: storage classes (Standard, Nearline, Coldline, Archive — different access frequencies), object versioning, lifecycle rules for cost management, signed URLs for time-limited access, uniform vs. fine-grained bucket IAM.
Cloud SQL vs. Cloud Spanner: Cloud SQL is managed PostgreSQL/MySQL — regional, familiar, appropriate for most transactional workloads. Cloud Spanner is Google's globally distributed, horizontally scalable relational database — globally consistent, handles millions of QPS, expensive. Choose Spanner when you need global consistency at scale. Most workloads don't justify Spanner's cost.
Bigtable: NoSQL wide-column store, inspired by Google's internal Bigtable paper. Designed for high-throughput, low-latency reads/writes at petabyte scale. Best for time-series data, IoT telemetry, and analytics workloads where you have a clear row key design. Row key design is critical — poor key design causes hotspots.
Firestore: Document database with real-time sync. Better for mobile/web apps needing offline support and real-time listeners. Not for high-volume analytics.
Data: BigQuery
BigQuery is GCP's signature service and the most common interview topic for data engineering roles on GCP.
Key concepts: serverless analytics, columnar storage, separation of storage and compute, slot-based pricing vs. on-demand pricing, partitioned and clustered tables (partition on date/timestamp for cost and performance, cluster on high-cardinality columns used in WHERE clauses).
Interview question: "How do you optimize a slow BigQuery query?" Answer framework: check if you're scanning full tables (add partitioning), check column selection (avoid SELECT *, columnar storage rewards selecting only needed columns), use clustering on join/filter columns, push filters as early as possible, use APPROXCOUNTDISTINCT for large cardinality counts.
Messaging: Pub/Sub
Cloud Pub/Sub is GCP's managed message queue / event streaming service. Key properties: at-least-once delivery, push or pull subscription, global message ordering (with ordering keys), exactly-once processing (with deduplication windows on Pub/Sub Lite).
Interview: "Design a pipeline that processes events from Pub/Sub into BigQuery." Answer: Pub/Sub → Cloud Dataflow (Apache Beam) → BigQuery streaming inserts, or Pub/Sub → BigQuery direct subscription (now GA). Handle late data in Dataflow using watermarks and windowing.
IAM and Security
GCP IAM: principals (users, service accounts, groups), roles (primitive, predefined, custom), policy bindings. The principle of least privilege is tested: use predefined roles when possible, create custom roles only when predefined roles are too broad.
Service accounts: the GCP-native way for services to authenticate. Best practice: each service gets its own service account with minimal permissions. Avoid using the default Compute Engine service account — it has Editor permissions project-wide.
Organization hierarchy: Organization → Folders → Projects → Resources. IAM policies are inherited downward. This is a common interview topic for enterprise GCP: how do you enforce security policies across a large organization? Answer: Organization Policies (constraints), VPC Service Controls (data exfiltration prevention), centralized logging with Log Sinks to a security project.
Networking
VPC (Virtual Private Cloud) in GCP: global (unlike AWS where VPCs are regional), subnets are regional. Shared VPC for connecting multiple projects to a common network. VPC Peering for connecting two VPCs. Cloud Interconnect / Cloud VPN for on-premises connectivity.
Cloud Load Balancing: several types — Global External (HTTP/HTTPS, anycast IP), Regional External (TCP/UDP), Internal. GCP's global LB is unique: single anycast IP routes to the nearest healthy backend globally, providing both load balancing and global routing.
Interview Preparation Tips
GCP interviews at cloud engineering roles often include a whiteboard architecture problem. Common: "Design a data pipeline to ingest 1 million events/second from IoT devices." Expected answer: IoT Core (or direct Pub/Sub publish) → Pub/Sub → Dataflow → Bigtable (for real-time queries) + BigQuery (for analytics).
Know the GCP decision matrix: Cloud Run vs. GKE, Cloud SQL vs. Spanner, Bigtable vs. Firestore vs. BigQuery. Being able to articulate these tradeoffs clearly is what differentiates prepared candidates.
Ready to practice for your next interview? Interview Simulator gives you AI-powered feedback on your responses — behavioral, technical, and system design. Start with 3 free practice interviews today.
Try Interview Simulator free →
Don't neglect behavioral preparation — see our behavioral interview guide.
For technical interview preparation, our system design guide is essential reading.
Master the most common coding interview patterns to ace your technical rounds.
Your resume is the first impression — get it right with our resume guide.
Explore Related Topics
- Google Cloud Platform (GCP) BigQuery, Cloud Run, and GCP...
- AI Company Interview Guide 2026: OpenAI, Anthropic, and...
- AMD Software Engineer Interview Guide