120+ Cloud Computing Interview Questions & Answers: The Master Review

Q1. Explain the differences between Scalability and Elasticity with real-world examples.

Answer: Scalability is the architectural capability of a system to handle growing amounts of work by adding resources (e.g., upgrading from a 4-core database to a 16-core database over months of steady user growth). Elasticity is the system's ability to automatically scale resources up and down in real-time in response to dynamic demand fluctuations (e.g., an e-commerce platform automatically spinning up 50 extra web servers during a 2-hour Black Friday flash sale and shutting them down immediately when the sale ends).

Q2. What is the difference between Synchronous and Asynchronous replication in cloud databases?

Answer: In Synchronous replication (used in Multi-AZ RDS deployments), a database write transaction is not committed until data is written to both the primary instance AND the standby replica simultaneously. This guarantees zero data loss (RPO = 0) at the expense of slight write latency. In Asynchronous replication (used in Read Replicas and cross-region replication), the write is committed immediately on the primary, and replicated in the background to replicas with slight replication lag.

Q3. How do you design an application to achieve 99.999% ("Five Nines") availability in the cloud?

Answer: Five Nines allows only 5.26 minutes of total downtime per year. To achieve this: (1) Eliminate all single points of failure by deploying across multiple global regions in an Active-Active configuration with Route 53 latency/health check routing, (2) Use serverless or containerized microservices that auto-scale and self-heal in seconds, (3) Use distributed multi-region databases with active master replication (e.g., DynamoDB Global Tables or Google Cloud Spanner), and (4) Implement blue/green zero-downtime deployment pipelines with automated rollback gates.

Q4. How does a NAT Gateway differ from an Internet Gateway in AWS VPC?

Answer: An Internet Gateway (IGW) enables two-way communication: instances with public IPs in a public subnet can receive incoming connections from the internet and initiate outgoing connections. A NAT Gateway enables one-way communication: instances with only private IPs in a private subnet can initiate outbound connections to the internet (e.g., to fetch software updates), but outside internet users are completely blocked from initiating inbound connections to those instances.

Q5. What is the difference between AWS IAM Roles and IAM Users? When should each be used?

Answer: An IAM User has permanent credentials (username/password or long-term Access Key ID/Secret Access Key) and should only be created for individual human administrators (with mandatory MFA). An IAM Role has no long-term credentials; instead, it is assumed dynamically by trusted entities (like an EC2 instance, Lambda function, or GitHub Actions runner) which receive short-lived, automatically expiring security tokens via AWS STS.

Q6. What happens internally when you execute 'kubectl run my-app --image=nginx'?

Answer: (1) kubectl parses the command and issues an HTTP POST request to kube-apiserver. (2) The API Server authenticates and authorizes the user, validates the manifest schema, and persists the new Pod object to etcd. (3) kube-scheduler detects a pending Pod with no assigned node, evaluates node filters/scoring, and assigns a healthy worker node to the Pod. (4) The kubelet agent on that specific worker node detects the assignment, commands the container runtime (containerd) to pull the NGINX image and spawn the container, and sets up network namespaces via CNI plugins. (5) Kubelet updates the API server that the Pod is Running.

Q7. How do you mitigate the Serverless Cold Start problem for latency-sensitive APIs?

Answer: (1) Choose compiled or lightweight interpreted runtimes (Node.js, Python, Go, or Rust) instead of heavyweight JVM/CLR runtimes, (2) Configure Provisioned Concurrency in AWS Lambda to keep a pool of execution environments pre-warmed 24/7, (3) Optimize bundle size using tools like ESBuild to remove unnecessary npm dependencies, (4) Increase allocated memory (which proportionally allocates more CPU power to accelerate initialization), and (5) Initialize database connection pools outside the handler function for warm container reuse.

Q1. How should I prepare for a Cloud Solutions Architect interview?

Q2. What certifications carry the highest value in the cloud industry?