Perfect. Below is a pure interview-only AWS service selection question bank (120 questions) — no tutorials, no fluff, exactly the kind of “which service & why” questions used in real interviews for Data Engineer / Python / PySpark roles.

Use this as:

  • 🔹 mock interview practice
  • 🔹 rapid revision
  • 🔹 service-selection clarity

🎯 AWS SERVICE SELECTION — INTERVIEW QUESTION BANK (120)

🧮 COMPUTE (EC2 · Lambda · ECS · EMR · Glue)

  1. When would you choose EC2 over Lambda?
  2. When does Lambda become a bad choice?
  3. Why is Lambda not suitable for Spark?
  4. EC2 vs ECS — what problem does ECS solve?
  5. When would you choose containers instead of EC2?
  6. Why is EMR preferred over EC2 for Spark?
  7. Glue vs EMR — which one gives more Spark control?
  8. When should you avoid Glue Spark jobs?
  9. Why is EMR Serverless better than classic EMR in some cases?
  10. EMR vs Databricks — selection criteria?
  11. When would you use multiple EC2 instances instead of EMR?
  12. What compute service suits long-running Python services?
  13. What service is best for event-driven execution?
  14. Can Lambda trigger EMR jobs? Why would you do that?
  15. Why is Glue considered an ETL service, not general compute?
  16. When does ECS make more sense than Lambda?
  17. Why would you NOT use EMR Serverless?
  18. How does compute choice affect AWS cost?
  19. Why do companies still use EC2 despite serverless?
  20. What compute service fits batch vs streaming Spark?

🗄️ STORAGE (S3 · EBS · EFS · Glacier)

  1. Why is S3 not a file system?
  2. When would you choose EBS over S3?
  3. EFS vs S3 — key difference?
  4. Why is S3 preferred for data lakes?
  5. Why is rename expensive in S3?
  6. What happens if you store Spark shuffle data in S3?
  7. When would you use Glacier?
  8. S3 vs HDFS — which one is cheaper at scale?
  9. Why is S3 ideal for ephemeral EMR clusters?
  10. Why is EFS rarely used in data engineering?
  11. How does storage choice affect Spark performance?
  12. Can Lambda write to EBS?
  13. When would you attach multiple EBS volumes?
  14. Why does Athena require S3?
  15. Why do Glue crawlers need S3 paths?
  16. Why is S3 strongly consistent important?
  17. When should you avoid too many S3 partitions?
  18. Why is S3 used instead of databases for analytics?
  19. How does lifecycle policy reduce cost?
  20. Why is S3 better than EFS for large datasets?

🧾 DATABASES (RDS · DynamoDB · Glue Catalog)

  1. When would you choose RDS over DynamoDB?
  2. Why is DynamoDB bad for joins?
  3. Why is Glue Catalog NOT a database?
  4. Why does Spark not write data into Glue Catalog?
  5. When is Athena preferred over RDS?
  6. Can Glue Catalog replace RDS? Why not?
  7. Why do OLTP workloads avoid S3?
  8. Why is DynamoDB serverless?
  9. Why is RDS not ideal for analytics?
  10. What database fits metadata storage?
  11. Why is schema evolution easier in Glue?
  12. Why is DynamoDB not used as a data lake?
  13. When would you store small reference data in RDS?
  14. Why does Athena not support updates?
  15. How does database choice affect ETL design?

🔁 ORCHESTRATION (Step Functions · Airflow · EventBridge)

  1. Step Functions vs Airflow — when to choose which?
  2. Why is Step Functions called “serverless Airflow-lite”?
  3. Why is Airflow still used despite Step Functions?
  4. Can Step Functions replace Airflow completely?
  5. When would you use EventBridge instead of Step Functions?
  6. Why is Lambda often combined with Step Functions?
  7. Why should orchestration be decoupled from compute?
  8. Why is retry logic important in workflows?
  9. Why does Glue integrate well with Step Functions?
  10. When would you avoid Airflow?
  11. Can Step Functions trigger EMR jobs?
  12. Why is EventBridge good for loosely coupled systems?
  13. Why are DAGs better than scripts?
  14. What happens if orchestration fails?
  15. Why is orchestration a senior-level topic?

🧪 CI/CD (CodePipeline · GitHub · CloudFormation)

  1. CI/CD vs Infrastructure as Code — difference?
  2. Why is CloudFormation not CI/CD?
  3. Why do many teams use GitHub instead of CodeCommit?
  4. When would you choose CodePipeline?
  5. Why should infrastructure be version-controlled?
  6. Can CloudFormation deploy Spark jobs?
  7. Why is CI/CD important for data pipelines?
  8. How does CI/CD help rollback?
  9. Why is CI/CD rarely used for Glue scripts in small teams?
  10. What breaks if CI/CD is missing?

🏗️ INFRASTRUCTURE (CloudFormation · Terraform)

  1. CloudFormation vs Terraform — when to choose which?
  2. Why is IaC critical in production?
  3. Can CloudFormation create EMR clusters?
  4. Why is manual console creation risky?
  5. Why do companies prefer Terraform for multi-cloud?
  6. Why is IaC important for cost control?
  7. Why should IAM roles be created via IaC?
  8. Can IaC manage Glue jobs?
  9. Why does IaC reduce human error?
  10. Why is IaC expected in senior roles?

📊 MONITORING & SECURITY (CloudWatch · IAM · CloudTrail)

  1. Why is CloudWatch mandatory in production?
  2. Why is IAM the most common failure point?
  3. Why should services use IAM roles instead of keys?
  4. What happens if IAM permissions are too broad?
  5. Why is CloudTrail important for audits?
  6. Why should Lambda always log to CloudWatch?
  7. How does monitoring help cost optimization?
  8. Why is security shared responsibility?
  9. Why is least privilege important?
  10. Why is IAM tested heavily in interviews?

🔥 SCENARIO-BASED (REAL INTERVIEW QUESTIONS)

  1. You need to process 5 TB daily batch data — what services?
  2. You need near real-time ingestion — which services?
  3. You want zero server management — what do you choose?
  4. You need SQL-only access for analysts — which tool?
  5. You want cheapest long-term storage — what service?
  6. You need retries + branching logic — which orchestration?
  7. You want fast Spark startup — which option?
  8. You need cross-account access — how?
  9. You want audit trail of user actions — what service?
  10. You want to avoid vendor lock-in — what IaC?
  11. You want fully serverless ETL — which stack?
  12. You need containerized Spark — what service?
  13. You want automatic schema detection — which service?
  14. You need metadata only — what service?
  15. You want to minimize AWS bill — what design choice?
  16. You want easy rollback — which practice?
  17. You need Spark streaming — which compute?
  18. You want no cluster lifecycle — which service?
  19. You want tight Spark tuning — which option?
  20. You want simplest architecture — which services?

🧠 HOW TO USE THIS EFFECTIVELY

  • ✅ Answer verbally (no notes)
  • ✅ For each question, say:
    Service → Why → Why not others
  • ✅ Practice 10/day → 12 days = mastery

🚀 NEXT (Highly Recommended)

Reply with one or more:

  • A → Model answers for all 120
  • B → Mock interview (I ask, you answer)
  • C → Turn this into a printable PDF
  • D → AWS + PySpark scenario-based coding interview

You’re now thinking at senior AWS engineer level — this is exactly how interviewers think 💯

Pages: 1 2 3 4