Cloud Architecture Interview Questions for India 2026

Cloud architecture interviews test design thinking more than service memorisation. Core concepts: High availability vs fault tolerance: HA means minimising downtime (achieved through redundancy); fault tolerance means zero downtime even during failures (more expensive, usually reserved for critical systems). RPO and RTO: Recovery Point Objective (how much data loss is acceptable) and Recovery Time Objective (how long the system can be down). These drive disaster recovery design. Multi-region vs multi-AZ: multi-AZ handles data centre failures (availability zones within one region). Multi-region handles regional failures or global latency requirements. Circuit breaker pattern: prevents cascade failures in microservices: stops calls to a failing service and returns a fallback. CAP theorem: distributed systems can guarantee only two of Consistency, Availability, and Partition Tolerance.

Common AWS architecture design questions at Indian interviews: Design a highly available web application on AWS. Answer: Route 53 (DNS + health checks), CloudFront (CDN), Application Load Balancer, Auto Scaling Group of EC2 instances in multiple AZs, RDS Multi-AZ (PostgreSQL), ElastiCache Redis (session and cache), S3 (static assets), CloudWatch (monitoring). Design a real-time notification system. Answer: API Gateway + Lambda (ingestion), Kinesis Data Streams (real-time processing), Lambda consumers, DynamoDB (state storage), SNS/SQS (fan-out to email, push, SMS). Design a file processing pipeline. Answer: S3 (upload trigger), SQS queue (decoupled processing), Lambda or ECS tasks (processing workers), DynamoDB (job status tracking), S3 (output storage), CloudFront (download delivery).

Serverless (Lambda/Cloud Functions) vs containers (ECS/GKE): Serverless strengths: zero infrastructure management, automatic scaling, pay-per-execution (cost-efficient at low volume), fast deployment. Serverless limitations: cold start latency (critical for customer-facing, latency-sensitive APIs), maximum execution time (Lambda = 15 minutes), difficult to debug, stateless by design. Containers (ECS/EKS/GKE) strengths: predictable performance, long-running processes, stateful workloads, full control over environment. When to choose: event-driven, intermittent workloads: serverless. Consistent high-traffic APIs: containers. ECS vs EKS: ECS is AWS-native and simpler to manage for most use cases. EKS is Kubernetes: preferred when team already has k8s expertise or needs vendor-agnostic portability.

Cost optimisation is increasingly tested at senior cloud roles: Right-sizing: identify underutilised instances with CloudWatch metrics. EC2 instance types: On-Demand (full price, no commitment), Reserved Instances (1-3 year commitment, 30-60% discount), Spot Instances (spare capacity, 70-90% discount, can be interrupted: suitable for batch jobs, ML training). S3 storage tiers: Standard, Infrequent Access, Glacier: lifecycle policies to move data to cheaper tiers automatically. Data transfer costs: often overlooked: egress from AWS is charged, cross-AZ data transfer is also charged. Cloud security: IAM (principle of least privilege, role-based access, no long-lived access keys), VPC security groups vs NACLs, encryption at rest (KMS) and in transit (TLS), AWS Config and CloudTrail for compliance and audit.