AWS Interview Questions with a Shared-Zone Dependency Case
Strong AWS interview answers connect a service to a requirement, a failure assumption and a check. These original practice questions cover compute, storage, databases, networking, security, operations and migration. A worked two-zone example shows why separate instances do not prove independent availability when both use one zone's NAT gateway.
Topic: Cloud Computing: AWS, Azure & GCP in IT Infrastructure · Interview preparation
Quick answer: Strong AWS interview answers connect a service to a requirement, a failure assumption and a check. These original practice questions cover compute, storage, databases, networking, security, operations and migration. A worked two-zone example shows why separate instances do not prove independent availability when both use one zone's NAT gateway. Use the answer cues to explain trade-offs, and check current service availability before recommending an older product. The scenario uses supplied records and does not claim a deployed AWS test.
Key takeaways
- Trace dependencies before claiming multi-zone resilience.
- Distinguish RDS deployment types and S3 storage classes.
- Use temporary credentials and appropriate least-privilege access.
- Check current eligibility for older AWS services.
- Tie cost estimates to dated usage assumptions.
Table of Contents
- How should you answer AWS compute and scaling questions?
- Why can a two-zone design still fail together?
- How do storage and database answers change with requirements?
- What should networking and security answers establish?
- How do you distinguish operational and integration services?
- Which older migration and search answers need qualification?
- How can you turn these answers into useful interview practice?
- References
How should you answer AWS compute and scaling questions?
1. What is AWS? A collection of cloud services with different operational and billing models. Start with the workload requirement before selecting a product. 2. EC2 or Lambda? EC2 provides virtual machines whose operating systems you manage; Lambda executes functions in a managed service with runtime, duration and integration constraints. 3. What are AMIs and Auto Scaling? An AMI supplies an instance launch image. Auto Scaling adjusts a configured group; useful scaling still depends on health signals, limits and application behaviour. 4. ECS, EKS or Elastic Beanstalk? ECS orchestrates containers, EKS provides managed Kubernetes, and Beanstalk manages application deployment infrastructure. Each leaves customer responsibilities, including application security and suitable configuration. 5. What does a load balancer solve? It distributes supported traffic to eligible targets and uses configured health checks. It cannot repair an unhealthy application or prove a business transaction succeeded.
Why can a two-zone design still fail together?
6. Does placing two instances in different Availability Zones prove resilience? No. Trace the dependencies required by the actual request. In this fictional design, instances in zones A and B both call an external verification service through one NAT gateway in A. The NAT gateway guide explains the resulting shared-zone dependency. 7. What should the candidate propose? Review zone-local egress and routing, then check other shared dependencies and the cost of the proposed change. A diagram change needs an approved failure test and application acceptance criteria. 8. How does this differ from disaster recovery? Multi-zone availability addresses selected failures within a Region. Regional recovery also needs usable data, configuration, dependencies and a tested recovery procedure.
| Supplied observation | Reasoning | Next check |
|---|---|---|
| Application instances occupy A and B | Compute placement spans zones | Confirm traffic and health-check routing |
| Both private subnets use the NAT gateway in A | External calls share a zone dependency | Inspect supplied route associations |
| Zone A becomes unavailable | Instance B may lose required external access | Validate the full request path in an approved exercise |
Explain the conditional result precisely: loss of that egress dependency can break this application path, even if instance B remains running. It does not mean every private-subnet operation stops. Identify which calls need the external service and whether an approved alternative exists.
How do storage and database answers change with requirements?
9. S3 or EBS? S3 stores objects accessed through service interfaces; EBS provides block volumes for supported compute use. Object keys and mounted filesystem paths have different semantics. 10. Are all S3 classes multi-zone? No. The S3 storage-class guide includes single-zone options. Compare availability needs, access frequency, retrieval behaviour and minimum charging conditions. 11. What is an EBS snapshot? A point-in-time volume backup mechanism; application consistency and restore validation still require planning. 12. Multi-AZ or a read replica? Specify the deployment. The RDS Multi-AZ guide distinguishes a single-standby instance deployment from a two-standby cluster; their read-serving behaviour differs. A read replica is not a universal substitute for the required failover design. 13. Aurora, DynamoDB, Redshift, Neptune or ElastiCache? Match relational transactions, key-value/document access, analytical warehousing, graph queries or caching to the access pattern. A cache needs a safe source-of-truth and failure strategy.
What should networking and security answers establish?
14. What does a VPC provide? A logically isolated networking environment with subnets, routing and security controls. A subnet name alone does not establish public reachability. 15. Security groups, network ACLs and WAF? They operate at different scopes and layers; AWS WAF evaluates supported web requests, rather than replacing every network or application control. 16. IAM user or role? Explain the identity and required action. The IAM best-practice guide recommends temporary credentials and workload roles where appropriate, together with least privilege. 17. Do SSE and KMS grant access? Encryption and key management address protection of data and cryptographic operations. They do not replace application authorization or all resource permissions. 18. What do Secrets Manager and Organizations add? Managed secret handling and account governance respectively. Neither makes every consumer configuration safe automatically.
How do you distinguish operational and integration services?
19. CloudWatch or CloudTrail? CloudWatch supports operational telemetry and alarms; CloudTrail records supported AWS activity. Confirm selected event types and retention instead of assuming every data operation is recorded. 20. SNS, Kinesis, API Gateway or Step Functions? Distinguish notifications, streaming records, API management and workflow orchestration. Discuss delivery, retries, ordering and failure handling for the chosen service. 21. Route 53, CloudFront or Direct Connect? They address DNS, content delivery and dedicated network connectivity. DNS caching and application dependencies still affect recovery. 22. CloudFormation or CodePipeline? Infrastructure declarations and delivery orchestration solve different problems. A successful pipeline needs meaningful checks, not just a green aggregate. 23. Glue or EMR? Choose according to data integration and processing requirements, supported engines, operational control and team skills. CloudWatch Logs can help diagnose a job but does not validate its analytical result.
Which older migration and search answers need qualification?
24. Can a new customer simply order Snowball Edge? The current Snowball availability notice says it is unavailable to new customers and describes alternatives such as DataSync and Data Transfer Terminal. The main Snowball notice also announces support ending on 31 December 2026 in commercial Regions for Storage Optimized and Compute Optimized devices, with an exception for GovCloud and ADC customers holding active Snow jobs. Compare transfer size, connectivity, access and validation requirements. 25. Is CloudSearch a normal new-project option? The CloudSearch developer guide says it is closed to new customers. Existing-customer availability is different from a recommendation for a new account. 26. What do DMS and Storage Gateway do? DMS supports database migration and replication scenarios; Storage Gateway connects supported on-premises storage workflows to AWS. Compatibility, consistency and cutover checks determine whether they fit. 27. Does managed migration guarantee zero downtime? No. Source behaviour, replication lag, application cutover and rollback requirements must be evaluated.
| Interview requirement | Useful answer cue |
|---|---|
| Recover archived objects occasionally | Compare storage class and retrieval conditions |
| Cut over a transactional database | Validate consistency, lag and application acceptance |
| Estimate monthly cost | List resources, usage, Region and pricing assumptions |
| Prove recovery readiness | Restore and check the required service, not just backup existence |
How can you turn these answers into useful interview practice?
28. How should you estimate AWS cost? List compute duration, storage class, requests, transfer, supporting services and Region. Use current official pricing for the specified configuration instead of quoting an unexplained monthly number. 29. How do you answer an unfamiliar service question? State the requirement you understand, distinguish it from adjacent services and identify the official documentation needed to resolve the remaining choice. 30. What makes an experienced answer stronger? Follow a request from client to dependency, identify a failure assumption, and explain a check that could disprove it. For the two-zone example, redraw the egress dependency before claiming resilience. In a rehearsal, ask a partner to change one condition, such as removing the external call, and revise the conclusion accordingly.
Related reading: Cloud disaster-recovery interview cases; AWS audit-event coverage.
References
Frequently Asked Questions
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Wow
0
Sad
0
Angry
0