Cloud and DevOps fundamentals for campus placements — containers, CI/CD, and monitoring.
On-demand compute, storage, and managed services over the internet — pay as you go, scale quickly, no datacenter ops. TCS, Infosys, startups, and product firms deploy on AWS/Azure/GCP. Freshers should know IaaS vs PaaS vs SaaS at a high level.
IaaS: rent VMs, networks, disks (EC2). PaaS: deploy code without managing OS (Heroku, App Engine). SaaS: use software via browser (Gmail, Salesforce). You manage less stack as you move IaaS → SaaS.
Docker packages app + dependencies into immutable images run as containers on any host with consistent behavior. Solves 'works on my machine.' Enables microservices, faster CI, and Kubernetes scheduling. Know Dockerfile basics: FROM, COPY, RUN, EXPOSE, CMD.
VMs virtualize hardware with full guest OS — heavier isolation. Containers share host kernel, lighter and faster to start. VMs for strong multi-tenant isolation; containers for app density and DevOps velocity. Many systems use both.
Orchestrator scheduling containers across a cluster — self-healing, rolling updates, service discovery, autoscaling. Objects: Pod, Deployment, Service, Ingress. Fresher interviews rarely expect kubectl mastery but expect why orchestration matters at scale.
Trigger on push → checkout → install deps → lint → unit tests → build artifact (JAR/Docker image) → security scan → deploy to staging → smoke tests → promote to production (manual or automated). Tools: GitHub Actions, Jenkins, GitLab CI — common in campus hiring for DevOps-aware dev roles.
Define servers, networks, IAM in versioned files (Terraform, CloudFormation) instead of manual clicks. Enables reproducible environments, review in PRs, and disaster recovery. Mention drift detection and state files stored securely.
Dev: fast iteration, fake data. Staging: mirrors prod for integration/UAT. Production: real users, strict change control. Never test destructive scripts on prod; secrets differ per environment via env vars or secret managers.
Never commit .env with prod keys. Use AWS Secrets Manager, Azure Key Vault, HashiCorp Vault, or CI masked variables. Rotate credentials, least-privilege IAM roles for apps (not long-lived root keys).
Automatically add/remove instances based on CPU, memory, request rate, or queue depth. Handles Diwali sale or exam-result traffic spikes. Pair with load balancer health checks; scale-down carefully to avoid flapping.
Run two identical environments; switch traffic from blue (current) to green (new) atomically. Fast rollback by switching back. Reduces downtime vs in-place deploys. Canary releases send small % traffic first — even safer.
Logs: event details. Metrics: aggregated numbers over time. Distributed traces: follow one request across microservices (OpenTelemetry). Together debug latency in production — expected vocabulary for SRE-leaning campus roles.
EC2 (VMs), S3 (object storage), RDS (managed SQL), Lambda (serverless functions), IAM (permissions), VPC (networking), CloudWatch (monitoring), ECS/EKS (containers). Relate to a project: 'Static site on S3 + CloudFront, API on EC2 or Lambda.'
Run functions without managing servers — cloud scales execution per invocation (AWS Lambda). Pay per use; cold starts matter. Good for event-driven tasks (image resize, webhooks). Not for long-running heavy compute without tuning.
Reverse proxy sits in front of servers (NGINX, ALB) — SSL, routing, caching. Forward proxy sits in front of clients (corporate internet gateway). DevOps interviews mention NGINX terminating HTTPS and load balancing to Node/Java backends.
Git as single source of truth for desired infrastructure/app state; controllers reconcile cluster to match repo. Argo CD and Flux popular with Kubernetes. Shows modern deployment thinking beyond manual kubectl apply.
Commands: grep, tail, chmod, systemd services, curl, ssh, top/htop, file permissions, cron. Debugging production often starts on a shell — campus candidates with Linux lab experience have an edge.
SLI: measured indicator (successful requests / total). SLO: target (99.9% monthly). Error budget: allowed unreliability before feature freeze to focus on stability. Language from Google SRE book — increasingly quoted in Indian product interviews.
Use one stage to compile (Maven/Go build) and a slim runtime stage copying only artifacts — smaller, more secure images. Demonstrates production-minded container practices in project discussions.
Frontend on Vercel/Netlify or S3+CloudFront; backend API in Docker on ECS/Render; PostgreSQL on RDS; env secrets in manager; CI runs tests on PR; HTTPS everywhere. Concrete story beats listing buzzwords.
