Cloud Interview Prep (AWS, GCP, Azure)

AWS interview questions are unavoidable in any cloud-leaning backend interview — the core compute services (EC2 vs ECS vs EKS vs Fargate vs Lambda), RDS and Aurora and DynamoDB, S3 and storage classes, IAM (roles vs users, least privilege, the trust policy that always confuses), the messaging stack (SQS vs SNS vs EventBridge vs Kinesis), serverless patterns with Lambda + Spring Cloud Function, and the production-debug scenarios interviewers actually ask ("the AWS bill doubled overnight — debug it"). For vendor-agnostic cloud-native patterns (12-factor, deployment strategies, DR) see the Cloud Native page; for GCP and Azure see the dedicated platform pages.

Pick the smallest AWS service that does the job.

• Compute — Lambda < ECS Fargate < ECS / EKS < EC2.
• Datastore — DynamoDB for KV, RDS for relational, S3 for blobs.
• Messaging — SQS for queues, SNS for fan-out, EventBridge for rules, Kinesis for streams.
• Identity — IAM roles for everything; never long-lived keys.
• Networking — VPC, subnets, security groups, NACLs.

The interview surface for this module breaks into these focus areas:

• Aws core services — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Aws messaging — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Comparisons — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Ecs and fargate — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Iam and security — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Rds with spring — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• S3 storage — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Serverless — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Scenario-based — applied questions tying multiple topics together. Drill these last, after the fundamentals are sharp.

• IAM least-privilege — explicit deny > implicit; trust policies trip people up.
• S3 consistency — strong read-after-write since 2020; pre-2020 trivia is outdated.
• DynamoDB partition key is permanent; pick to avoid hot partitions.
• Lambda cold starts — provisioned concurrency or SnapStart for Java.
• Cost — most expensive thing you forgot to turn off.

"I'd run the service on behind , store data in picked for , use SQS for buffering, attach IAM roles per-service, and put cost guardrails on day one."

Google Cloud Platform (GCP) interview questions cover the platform powering an increasing share of Java backend roles — the core GCP services for backend developers (Cloud Run, GKE, Cloud SQL, Spanner, BigQuery), Pub/Sub messaging, Cloud Storage, IAM and Workload Identity, Secret Manager, Cloud Monitoring, and the deployment patterns for Spring Boot on Cloud Run vs GKE. Use this page when the job description specifies GCP — the questions are different enough from AWS that a generic cloud answer will not cut it.

GCP = opinionated, managed-first Google Cloud.

• Compute — Cloud Run < GKE < GCE.
• Data — Cloud SQL (relational), Spanner (global SQL), BigQuery (analytics), Firestore (KV/doc).
• Messaging — Pub/Sub for everything async.
• IAM — workload identity > service-account keys.
• Operations — Cloud Monitoring + Logging + Trace, all integrated.

The interview surface for this module breaks into these focus areas:

• Gcp compute and deployment — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Gcp core services — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Gcp messaging — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Gcp observability — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Gcp security and iam — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Gcp storage and data — drill the named patterns, the failure modes, and the recommended Java / framework idiom.

• Cloud Run autoscales to zero — cheap, but cold starts.
• Spanner is the only globally-consistent SQL — at a price.
• Workload identity — no static keys for GKE workloads.
• Service-account JSON keys are the security smell — rotate or eliminate.
• BigQuery on-demand vs flat-rate — pick by predictability of spend.

"For Spring Boot on GCP I'd ship Cloud Run with a min-instance for warmth, use Cloud SQL for transactional data, BigQuery for analytics, secrets in Secret Manager, identity via Workload Identity, and observability via Cloud Operations."

Microsoft Azure interview questions cover the platform common in enterprise Java backend roles — Azure Spring Apps for managed Spring Boot, AKS and Azure Container Registry, Azure Functions for serverless Java, the storage stack (Blob, SQL Database, Cosmos DB), the messaging stack (Service Bus vs Event Hubs), Key Vault and Managed Identity for passwordless auth, Azure Active Directory with Spring Security, and Application Insights for observability. Use this page when the job description mentions Azure or Microsoft-shop tooling.

Azure = enterprise-first Microsoft cloud.

• Compute — Azure Spring Apps (managed Spring Boot), AKS, Functions.
• Data — Azure SQL, Cosmos DB (multi-model), Blob Storage.
• Messaging — Service Bus (queues / topics), Event Hubs (streams).
• Identity — Entra ID (formerly AAD); managed identities for services.
• Observability — Application Insights tied to Azure Monitor.

The interview surface for this module breaks into these focus areas:

• Azure api and observability — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Azure compute and deployment — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Azure core services — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Azure messaging — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Azure security and iam — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Azure storage and data — drill the named patterns, the failure modes, and the recommended Java / framework idiom.

• Cosmos DB partition key — hot partitions blow your RU bill; choose carefully.
• Managed Identity > connection strings.
• Key Vault + reference syntax in App Settings — pulls secrets at runtime.
• Spring on Azure — Azure Spring Apps gives you Eureka + Config without standing them up.
• Hybrid scenarios — Azure Arc / ExpressRoute show up in enterprise interviews.

"I'd deploy Spring Boot to Azure Spring Apps or AKS, store transactional data in Azure SQL, use Service Bus for async, secure with Managed Identity + Key Vault, and instrument with Application Insights for end-to-end traces."

Cloud-native interview questions cover the cloud-vendor-agnostic patterns every modern backend interview probes — the 12-factor methodology, cloud design patterns (circuit breaker, bulkhead, retry-with-backoff, ambassador, sidecar), deployment strategies (rolling, blue/green, canary, progressive delivery), disaster recovery (RPO/RTO), multi-region active/active vs active/passive trade-offs, cost optimization (FinOps), and the orchestrator-choice question (Kubernetes vs ECS vs Cloud Run vs Lambda). Pair this with the AWS, GCP, Azure or Kubernetes pages for the platform-specific surface.

Cloud-native ≠ cloud — it's the patterns that work in any cloud.

• 12-factor app — config in env, stateless processes, log to stdout, etc.
• Patterns — sidecar, ambassador, circuit breaker, bulkhead.
• Deployment — rolling, blue/green, canary, progressive delivery.
• DR — RPO (data loss) + RTO (downtime); design backwards from those numbers.
• FinOps — cost is a non-functional requirement.

The interview surface for this module breaks into these focus areas:

• 12 factor app — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Cloud design patterns — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Comparisons — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Cost optimization — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Deployment strategies — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Disaster recovery — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Multi region architecture — drill the named patterns, the failure modes, and the recommended Java / framework idiom.
• Scenario-based — applied questions tying multiple topics together. Drill these last, after the fundamentals are sharp.

• Active/active vs active/passive — latency vs cost vs complexity.
• Stateless ≠ no state — externalise state to managed stores.
• Multi-region — DNS-based failover or anycast; both cost.
• Cost optimisation — reservations + autoscaling + savings plans.
• Vendor lock-in is partly a myth — use managed services, abstract carefully.

"I'd build to 12-factor, run stateless services with externalised state, use canary deploys with auto-rollback, plan DR around explicit RPO/RTO, and bake cost dashboards into the SRE rituals."