Name: CO5 Cloud Computing - Detailed Notes
Uploaded: 2026-04-22
Duration: 19 min
Description: UNIT CO5. Cloud Computing. IAPIs · Pricing · Migration · Web Services · Big Data · Case Studies.

UNIT CO5. Cloud Computing. IAPIs · Pricing · Migration · Web Services · Big Data · Case Studies. 

Scene 1

IAPIs · Pricing · Migration · Web Services · Big Data · Case Studies

Section 1 → REST & API-Based Access (Slides 2–5) Section 2 → Cloud Pricing & Cost Optimization (Slides 6–9) Section 3 → Moving Applications to Cloud (Slides 10–13) Section 4 → Web Services & Cloud Technologies (Slides 14–17) Section 5 → Data-Intensive Technologies (Slides 18–21) Section 6 → Case Studies: Netflix · Flipkart · MeghRaj (Slides 22–28)

SECTION 1. REST & API-Based Access in Cloud. HTTP Methods · REST Principles · API Gateway · OAuth · SDK vs REST. 

Scene 2

HTTP Methods · REST Principles · API Gateway · OAuth · SDK vs REST

What is a REST API? — Deep Dive. Definition: REST (Representational State Transfer) is an architectural style for designing networked applications. It uses HTTP requests to perform CRUD operations on resources identified by URLs.. 

Scene 3

Definition: REST (Representational State Transfer) is an architectural style for designing networked applications. It uses HTTP requests to perform CRUD operations on resources identified by URLs.

Separation of UI (client) from data storage (server). Improves portability & scalability.

Each request must contain all info needed. Server stores no client session data.

Responses must define themselves as cacheable or non-cacheable to improve efficiency.

Consistent interface between components: resource identification, manipulation via representations.

Client cannot tell whether connected to end server or intermediary. Enables load balancers, caches.

(Optional) Servers can send executable code to clients (e.g., JavaScript). Extends client functionality.

HTTP Methods & Status Codes in Cloud APIs. HTTP Methods (CRUD Operations):. 

Scene 4

HTTP Methods & Status Codes in Cloud APIs

API Gateway, Authentication & SDK vs REST. API Gateway — What & Why. 

Scene 5

API Gateway, Authentication & SDK vs REST

Definition:
 An API Gateway is a server that acts as the single entry point for all client requests to backend microservices.

 Functions of API Gateway:
 • Request routing to correct microservice
• Authentication & Authorization
• Rate limiting & throttling
• Load balancing
• SSL termination
• Logging & monitoring
• Response caching

 Examples: • AWS API Gateway • Azure API Management • Kong Gateway • Google Apigee

API Key: Simple token, sent in header/URL
 OAuth 2.0: Token-based, used by Google/Facebook APIs
 JWT: JSON Web Token, stateless auth
 Basic Auth: Username:Password (Base64 encoded)

REST API — Request & Response Structure. HTTP Request Structure. 

Scene 6

REST API — Request & Response Structure

1. Request Line:
    POST /api/v1/users HTTP/1.1

 2. Headers:
    Host: api.example.com
   Content-Type: application/json
   Authorization: Bearer <token>

 3. Body (for POST/PUT):
   

 4. Query Parameters:
    GET /users?role=admin&limit=10

1. Status Line:
    HTTP/1.1 201 Created

 2. Response Headers:
    Content-Type: application/json
   Cache-Control: no-cache
   X-RateLimit-Remaining: 99

 3. Response Body (JSON):
   

 Key Notes:
 • Always validate status code first • JSON is most common format • Use versioning: /api/v1/, /api/v2/

SECTION 2. Cloud Pricing Models & Cost Optimization. 

Scene 7

Pay-As-You-Go · Reserved · Spot · TCO · FinOps · Cost Tools

Cloud Pricing Models — In Detail. Pay-As-You-Go (PAYG). 

Scene 8

Charged per second/minute/hour of use. No minimum commitment. Ideal for unpredictable or new workloads. AWS charges EC2 by the second (min 60s). Highest per-unit cost but maximum flexibility.

1-year or 3-year commitment in exchange for 40-75% discount vs PAYG. Two payment options: All Upfront (cheapest) or Monthly. Best for steady, predictable workloads like databases, web servers.

Bid on unused cloud capacity. AWS Spot can be 70-90% cheaper than on-demand. Cloud provider can reclaim with 2-min notice. Use for batch jobs, CI/CD, ML training, video encoding.

Commit to a specific spend/hour (e.g., $10/hr) for 1-3 years. More flexible than RIs — applies across instance families and regions. GCP calls these Committed Use Discounts (CUDs).

AWS: 12 months free (EC2 t2.micro 750hrs, S3 5GB, RDS 750hrs). GCP: $300 credit for 90 days. Azure: 12 months free services + $200 credit. Ideal for learning, testing, small projects.

Used by SaaS providers. Fixed monthly fee regardless of usage. Examples: Microsoft 365, Google Workspace, Salesforce. Predictable billing — good for business planning.

Total Cost of Ownership (TCO) & Cost Optimization. 

Scene 9

Total Cost of Ownership (TCO) & Cost Optimization

TCO (Total Cost of Ownership): Compares on-premise infrastructure costs vs cloud. Includes hardware, power, cooling, space, staff, maintenance. AWS TCO Calculator and Azure Pricing Calculator help estimate migration savings.

Analyze CloudWatch/Azure Monitor metrics. Downsize over-provisioned instances. Use AWS Compute Optimizer for recommendations.

Scale out during peak, scale in during low traffic. Use HPA in Kubernetes. Set min/max limits. Prevents both waste and performance issues.

S3 Lifecycle policies: move to S3-IA after 30 days, Glacier after 90 days. Azure: Hot→Cool→Archive tiers. Reduce storage cost by 70%.

Use Spot/Preemptible for fault-tolerant jobs: ML training, video rendering, ETL pipelines. Combine with checkpointing to handle interruptions.

Analyze 3-month usage trends. Buy RIs for baseline, PAYG for spikes. AWS Savings Plans: flexible across instance types.

Financial Operations: shared responsibility for cloud costs. Use tags to track costs per team/project. Weekly cost reviews. Showback/chargeback to teams.

AWS vs Azure vs GCP — Pricing Comparison. Service. 

Scene 10

AWS vs Azure vs GCP — Pricing Comparison

Tip: Always use multi-cloud cost tools like CloudHealth, Apptio Cloudability, or native dashboards to avoid bill shock.

SECTION 3. Moving Applications to the Cloud. 6 R's Migration · Phases · Challenges · Hybrid Cloud · Containers. 

Scene 11

6 R's Migration · Phases · Challenges · Hybrid Cloud · Containers

The 6 R's of Cloud Migration — Deep Dive. Rehost — "Lift & Shift". 

Scene 12

The 6 R's of Cloud Migration — Deep Dive

Move app as-is using IaaS VMs. No code changes. Tools: AWS SMS, Azure Migrate, CloudEndure. Example: SAP on Azure VMs. Best when: speed of migration is priority.

Minor changes to leverage cloud benefits without redesigning. E.g., move MySQL to Amazon RDS (managed). Move app server to AWS Elastic Beanstalk. Less management overhead.

Redesign app using cloud-native patterns: microservices, serverless, containers. Example: Monolith → Kubernetes microservices. High upfront cost but maximum cloud benefit long-term.

Replace existing software with SaaS alternative. Email → Microsoft 365. CRM → Salesforce. HRMS → Workday. Eliminates infrastructure management entirely.

Keep on-premise: legacy mainframes, regulatory compliance systems, recently upgraded apps where migration ROI is low. Revisit in 12-18 months for potential migration.

Identify and shut down apps that are no longer used (often 10-20% of portfolio). Reduces licensing, maintenance, and operational costs immediately.

Cloud Migration Phases & Key Challenges. AWS Cloud Adoption Framework (CAF) Phases:. 

Scene 13

AWS Cloud Adoption Framework (CAF) Phases:

Inventory apps. Measure TCO. Define migration goals. Prioritize workloads.

Build cloud foundation. Set up landing zone. Train teams. Define governance.

Execute migration using 6 R's. Use AWS DMS/SMS, Azure Migrate tools.

Right-size resources. Enable auto-scaling. Implement cost management & monitoring.

Use encryption at rest & transit. Implement IAM. Check GDPR/ISO/SOC2 compliance.

Use Blue-Green deployments, canary releases. Migrate during off-peak hours.

Map all dependencies before migration. Use dependency discovery tools (AWS Application Discovery Service).

Train teams on cloud platforms. Use managed services to reduce ops burden. Hire cloud architects.

Set billing alerts. Use cost anomaly detection. Start with pilot workloads before full migration.

Conduct load testing post-migration. Use CDN. Optimize database queries. Enable caching.

Hybrid Cloud & Multi-Cloud Strategy. Hybrid Cloud. 

Scene 14

Definition:
 Combines on-premise (private cloud) with public cloud. Workloads can move between environments.

 When to Use Hybrid:
 • Regulated data must stay on-premise (banking, healthcare)
• Burst to cloud during peak loads (cloud bursting)
• Gradual migration strategy
• Legacy apps can't move to cloud

 Tools:
 • AWS Outposts (AWS hardware on-prem) • Azure Arc (manage on-prem via Azure) • Google Anthos (Kubernetes everywhere) • VMware Cloud on AWS

Definition:
 Using services from 2+ cloud providers simultaneously (e.g., AWS + GCP + Azure).

 Advantages:
 • Avoid vendor lock-in
• Use best service from each provider
• Improved redundancy & DR
• Negotiate better pricing
• Geographic coverage

 Challenges:
 • Increased complexity
• Skill requirements multiply
• Data transfer costs between clouds
• Unified monitoring is hard

 Management Tools:
 • Terraform (IaC across clouds) • Pulumi, Crossplane • Datadog, New Relic (monitoring)

SECTION 4. Cloud Technologies & Applications. SOAP vs REST · Containers · Serverless · Microservices · DevOps. 

Scene 15

SOAP vs REST · Containers · Serverless · Microservices · DevOps

SOAP vs REST — Complete Comparison. Criteria. SOAP. 

Scene 16

Containerization — Docker & Kubernetes. Docker — Container Platform. 

Scene 17

What is Docker?
 Platform to build, ship & run apps in containers. A container packages code + dependencies + config into a single portable unit.

 Key Concepts:
 • Image: Read-only blueprint (like a class)
• Container: Running instance of an image
• Dockerfile: Script to build an image
• Docker Hub: Registry for images
• Docker Compose: Multi-container apps

 VM vs Container:
 • VM: Full OS per app (GBs, slow boot)
• Container: Shared OS kernel (MBs, fast)
• Container starts in seconds vs VMs in minutes

 Cloud Container Services:
 • AWS ECS / EKS • Azure AKS • Google GKE

What is Kubernetes (K8s)?
 Open-source system (by Google) for automating deployment, scaling, and management of containerized applications.

 Key K8s Objects:
 • Pod: Smallest unit (1+ containers)
• Deployment: Manages pod replicas
• Service: Exposes pods to network
• Ingress: External traffic routing
• ConfigMap/Secret: Configuration
• Namespace: Logical isolation

 Features:
 • Auto-healing (restarts failed pods) • Horizontal Pod Autoscaler (HPA) • Rolling updates & rollbacks • Service discovery & load balancing • Storage orchestration

Serverless Computing & Microservices Architecture. 

Scene 18

Serverless Computing & Microservices Architecture

Definition:
 Run code without managing servers. Provider auto-scales, you pay per execution (not idle time).

 AWS Lambda (FaaS):
 • Triggered by events (S3 upload, API call, DynamoDB change)
• Supports Python, Node.js, Java, Go
• 15 min max execution time
• Free: 1M requests/month

 Use Cases:
 • Image/video processing on upload
• Real-time data transformation
• Chatbots & notification triggers
• Scheduled cron jobs

 Limitations:
 • Cold start latency • Stateless — no persistent memory • Vendor lock-in risk • Not for long-running processes

Definition:
 App built as a collection of small, independently deployable services. Each service handles one business function.

 Monolith vs Microservices:
 • Monolith: Single deployable unit, hard to scale specific parts
• Microservices: Each service scales independently

 Microservices Design Patterns:
 • API Gateway pattern (single entry point)
• Saga pattern (distributed transactions)
• CQRS (command-query separation)
• Circuit Breaker (fault tolerance)
• Service Mesh (Istio, Linkerd)

 Netflix Example:
 500+ microservices. Each team owns its service end-to-end. Independent deployments 100s of times per day.

SECTION 5. Data-Intensive Technologies for Cloud Computing. 

Scene 19

Data-Intensive Technologies for Cloud Computing

Hadoop · Spark · Kafka · NoSQL · Data Lakes · ML in Cloud

Apache Hadoop & Apache Spark — Big Data Engines. Apache Hadoop. 

Scene 20

Apache Hadoop & Apache Spark — Big Data Engines

Architecture:
 • HDFS (Hadoop Distributed File System): Stores data in blocks (128MB) across nodes. Replication factor=3 for fault tolerance.
• MapReduce: Programming model — Map (split/process) → Shuffle → Reduce (aggregate)
• YARN: Resource negotiator — schedules jobs across cluster

 MapReduce Word Count Example:
 Input: 'hello world hello'
Map: (hello,1),(world,1),(hello,1)
Shuffle: hello→[1,1], world→[1]
Reduce: hello=2, world=1

 Cloud Managed Hadoop:
 • AWS EMR (Elastic MapReduce) • Azure HDInsight • Google Cloud Dataproc

Why Spark over Hadoop?

 Spark processes data in-memory (RAM) instead of disk I/O. Up to 100x faster than MapReduce for iterative algorithms.

 Spark Components:
 • Spark Core: Base engine (RDDs)
• Spark SQL: SQL queries on big data
• Spark Streaming: Real-time data processing
• MLlib: Machine learning library
• GraphX: Graph computation

 Key Concept — RDD:
 Resilient Distributed Dataset: Fault-tolerant, immutable, distributed collection of objects that can be processed in parallel.

 Cloud:
 • AWS EMR with Spark, Azure Databricks (hosted Spark), Google Dataproc

Apache Kafka (Streaming) & NoSQL Databases. Apache Kafka — Event Streaming. 

Scene 21

Apache Kafka (Streaming) & NoSQL Databases

What is Kafka?
 Distributed event streaming platform. Handles millions of messages/second. Durable, fault-tolerant, high-throughput.

 Core Concepts:
 • Producer: Sends messages/events
• Consumer: Reads and processes events
• Topic: Named channel for messages
• Partition: Splits topic for parallelism
• Broker: Kafka server node
• ZooKeeper: Manages cluster metadata

 Use Cases:
 • Real-time analytics & fraud detection
• Activity tracking (LinkedIn clicks)
• Log aggregation from microservices
• Event sourcing in microservices

 Cloud:
 • Amazon MSK (Managed Streaming for Kafka) • Confluent Cloud • Azure Event Hubs (Kafka-compatible)

Simplest model. Key→Value. O(1) lookup. Ideal for caching, sessions, leaderboards.

Stores JSON/BSON documents. Flexible schema. Ideal for catalogs, user profiles, CMS.

Stores data in columns grouped into families. Optimized for write-heavy workloads, IoT data.

Nodes & relationships. Ideal for social networks, recommendation engines, fraud detection.

Data Lakes & Machine Learning in Cloud. Data Lake Architecture. 

Scene 22

Data Warehouse vs Data Lake:
 • Warehouse: Structured, schema-on-write, SQL, expensive (Redshift, Synapse)
• Lake: Raw/unstructured, schema-on-read, cheap object storage (S3, ADLS)

 Data Lake Zones:
 • Raw Zone: Ingest all data as-is
• Curated Zone: Cleaned & validated data
• Consumption Zone: Ready for analytics/ML

 AWS Lake Formation:
 • Builds data lakes on S3
• Integrates with Glue (ETL) + Athena (SQL query) + Redshift

 Azure Data Lake Storage (ADLS):
 • Hierarchical namespace over Blob • Integrates with Synapse Analytics + Databricks

AWS ML Services:
 • SageMaker: Build, train, deploy ML models
• Rekognition: Image & video analysis
• Comprehend: NLP & sentiment analysis
• Forecast: Time-series predictions

 Google Cloud AI:
 • Vertex AI: Unified ML platform
• Vision API, Speech API, NL API
• AutoML: No-code model training
• BigQuery ML: SQL-based ML

 Azure AI:
 • Azure ML Studio
• Cognitive Services (Vision, Speech, NLP)
• OpenAI Service (GPT-4, DALL-E)

 MLOps Lifecycle:
 Data → Feature Engineering → Training → Evaluation → Deployment → Monitoring

SECTION 6. Real-World Case Studies. Netflix · Amazon AWS · Flipkart · MeghRaj / NIC Cloud (Govt. of India). 

Scene 23

Netflix  ·  Amazon AWS  ·  Flipkart  ·  MeghRaj / NIC Cloud (Govt. of India)

Migrated to AWS. 500+ microservices, 15% internet traffic, global CDN.

Hybrid cloud for Big Billion Days. 100M+ users, 99.9% uptime maintained.

Govt of India cloud initiative. Hosts DigiLocker, AADHAAR, 600+ govt portals.

Case Study: Netflix — AWS Architecture Deep Dive. 

Scene 24

Case Study: Netflix — AWS Architecture Deep Dive

Context: Netflix serves 230M+ subscribers in 190+ countries. Streams 15% of global internet traffic. Migrated from own data centers to AWS between 2008-2016. Uses 100,000+ EC2 instances.

• EC2 Auto Scaling groups for transcoding servers • S3 stores 100PB+ of video content • CloudFront CDN distributes to 190+ countries • 200+ microservices handle different functions

• DynamoDB for user metadata & viewing history • Cassandra for billing & subscriber data • MySQL (RDS) for operational data • ElasticSearch for content search

• Apache Kafka for 700B+ events/day • Apache Spark for recommendation engine • S3 + EMR for analytics pipeline • Flink for real-time data processing

• Chaos Monkey: Randomly kills services in production to test resilience • Simian Army: Suite of chaos tools • Circuit Breaker pattern for fault tolerance • Active-Active multi-region deployment

• Spinnaker: Continuous delivery platform (open-sourced by Netflix) • Jenkins for CI pipelines • Docker + Titus (Netflix's container platform) • 100s of deployments per day with zero downtime

• Spot instances for batch transcoding (70%+ savings) • Reserved instances for databases • Auto-scaling reduces idle capacity • Built AWS Scryer for predictive scaling

Netflix — Chaos Engineering & Key Lessons. Chaos Engineering (Simian Army). 

Scene 25

Netflix — Chaos Engineering & Key Lessons

Philosophy: "Hope is not a strategy."
Deliberately break things in production to find weaknesses before they cause real outages.

 Simian Army Tools:
 • Chaos Monkey: Randomly terminates EC2 instances during business hours
• Chaos Gorilla: Terminates entire Availability Zone
• Chaos Kong: Simulates region failure
• Latency Monkey: Injects network delays
• Doctor Monkey: Health checks & removes unhealthy instances
• Security Monkey: Finds security vulnerabilities

 GameDay:
 Planned exercises where entire teams simulate disaster scenarios. Multiple AZ failure, database corruption, network partitions.

 Result:
 Netflix achieves 99.99% availability despite running 100% on cloud.

Key Learnings & Open Source Contributions

Open Source Tools by Netflix:
 • Spinnaker: Multi-cloud continuous delivery
• Hystrix: Circuit breaker for Java
• Eureka: Service discovery
• Zuul: API Gateway
• Ribbon: Client-side load balancing
• Conductor: Workflow orchestration

 Architecture Lessons:
 • Design for failure: assume every component will fail
• Stateless services scale horizontally
• Microservices enable independent deployments
• Multi-region active-active eliminates single points of failure
• Automate everything — manual processes don't scale

 Business Impact:
 • $1B+ savings vs on-premise estimate • 230M+ subscribers across 190 countries • 15% of global downstream internet traffic • Became the template for cloud-native architecture worldwide

CO5 — Netflix: Chaos Engineering & Lessons

Case Study: Flipkart — Cloud Migration & Big Billion Days. 

Scene 26

Case Study: Flipkart — Cloud Migration & Big Billion Days

Context: Flipkart (acquired by Walmart, 2018) is India's largest e-commerce platform. Handles 50M+ registered sellers, 150M+ products. Big Billion Days sale (Oct) sees 10x normal traffic in minutes.

Traditional data centers couldn't handle 10x traffic spikes during sales. Servers crashed in early sales (2014). Cart abandonment due to downtime cost ₹100s of crores. Manual capacity planning failed.

Adopted hybrid cloud: Own private cloud (Floppy) + Public cloud (Microsoft Azure). Microservices decomposition of monolith. Kubernetes for container orchestration. Service mesh for inter-service communication.

• Kafka for order event streaming • Cassandra for product catalog (billions of items) • Flink for real-time inventory updates • Redis for session management & cart • HBase for search index • Spark for analytics & recommendations

Load tests 3 months before BBD. Chaos testing in staging. Pre-warming caches. Pre-scaling Kubernetes clusters. Synthetic traffic simulation. Shadow mode testing with real prod traffic.

• Handled 1.5B+ page views/day • 100M+ app sessions in 24 hours • 99.9% uptime throughout 5-day sale • Orders delivered within 2 days • Zero major outages since 2015

• Read replicas for product catalog (DB reads >> writes) • Event-driven architecture decouples order → payment → fulfillment • Feature flags to gradually roll out sale features • Circuit breakers prevent cascade failures

Case Study: MeghRaj / NIC Cloud — Government of India. 

Scene 27

Case Study: MeghRaj / NIC Cloud — Government of India

Context: GI Cloud (MeghRaj) is India's national cloud initiative under Digital India programme, operated by NIC (National Informatics Centre) under Ministry of Electronics & IT. Announced 2013, operational since 2014.

Data Centers: NIC DCs in Delhi, Pune, Bhubaneswar, Hyderabad MPLS Network: High-speed connectivity between all NIC nodes 4000+ virtual servers deployed Tier 3+ data center standards Disaster Recovery: Primary-DR site replication Bandwidth: Multi-Gbps internet connectivity

IaaS: Virtual Machines, Storage, Networking PaaS: Databases, Web hosting, App engine SaaS: Email (NIC Mail), Video conferencing DigiLocker: 100M+ users, stores govt documents e-Office: Paperless government workflow GovCloud: Dedicated zones for sensitive data

AADHAAR: 1.3B biometric records backend National Scholarship Portal PFMS: Public Financial Management System e-Procurement portals (600+ ministries) GST Network initial hosting CoWIN: COVID vaccine management platform

CO5 — Complete Summary. Cloud APIs · Pricing · Migration · Web Services · Big Data · Case Studies. 

Scene 28

Cloud APIs · Pricing · Migration · Web Services · Big Data · Case Studies

6 Constraints · HTTP CRUD · Status codes · OAuth · API Gateway · SDK vs REST

PAYG · Reserved · Spot · Savings Plans · TCO · FinOps · AWS vs Azure vs GCP

6 R's · Rehost→Retire · CAF Phases · Challenges · Hybrid · Multi-Cloud

SOAP vs REST · Docker · Kubernetes · Serverless · Microservices · DevOps

Hadoop MapReduce · Spark RDD · Kafka streaming · NoSQL types · Data Lakes · Cloud ML

Netflix (chaos engineering, 200+ services) · Flipkart (BBD, hybrid) · MeghRaj (AADHAAR, DigiLocker)

CO5 Cloud Computing - Detailed Notes

Scene 1 (0s)

Scene 2 (18s)

Scene 3 (28s)

Scene 4 (1m 11s)

Scene 5 (2m 6s)

Scene 6 (2m 48s)

Scene 7 (3m 29s)

Scene 8 (3m 40s)

Scene 9 (4m 43s)

Scene 10 (5m 42s)

Scene 11 (6m 35s)

Scene 12 (6m 45s)

Scene 13 (7m 44s)

Scene 14 (8m 39s)

Scene 15 (9m 19s)

Scene 16 (9m 28s)

Scene 17 (10m 15s)

Scene 18 (11m 1s)

Scene 19 (11m 49s)

Scene 20 (12m 0s)

Scene 21 (12m 48s)

Scene 22 (13m 40s)

Scene 23 (14m 29s)

Scene 24 (14m 51s)

Scene 25 (15m 44s)

Scene 26 (16m 38s)

Scene 27 (17m 40s)

Scene 28 (18m 26s)